DESCRIPTION MOVING PICTURE CODING METHOD, AND MOVING PICTURE DECODING METHOD TECHNICAL FIELD The present invention relates to a moving picture coding method and a moving picture decoding method and, more particularly, to a method for coding or decoding pictures constituting a moving picture with reference to other pictures of the moving picture. BACKGROUND ART Generally, in coding of pictures constituting a moving picture, each picture is divided into plural blocks, and compressive coding (hereinafter, also referred to simply as "coding") of image information possessed by each picture is carried out for every block, utilizing redundancies in the space direction and time direction of the moving picture. As a coding process utilizing redundancy in the space direction, there is intra-picture coding utilizing correlation of pixel values in a picture. As a coding process utilizing redundancy in the time direction, there is inter-picture predictive coding utilizing correlation of pixel values between pictures. The inter-picture predictive coding is a process of coding a target picture to be coded, with reference to a picture that is positioned timewise forward the target picture (forward picture), or a picture that is positioned timewise backward the target picture (backward picture). The forward picture is a picture whose display time is earlier that that of the target picture, and it is positioned forward the target picture on a time axis indicating the display times of the respective pictures (hereinafter, referred to as "display time axis"). The backward picture is a picture whose display time is later than that of the target picture, and it is positioned backward the target picture on the display time axis. Further, in the following description, a picture to be referred to in coding the target picture is called a reference picture. In the inter-picture predictive coding, specifically, a motion vector of the target picture with respect to the reference picture is detected, and prediction data for image data of the target picture is obtained by motion compensation based on the motion vector. Then, redundancy of difference data between the prediction data and the image data of the target picture in the space direction of the picture is removed, thereby to perform compressive coding for the amount of data of the target picture. On the other hand, as a process for decoding a coded picture, there are intra-picture decoding corresponding to the intra- picture coding, and inter-picture decoding corresponding to the inter-picture coding. In the inter-picture decoding, the same picture as a picture that is referred to in the inter-picture coding is referred to. That is, a picture Xtg that is coded with reference to pictures Xra and Xrb is decoded with reference to the pictures Xra and Xrb. Figures 43(a)-43(c) are diagrams illustrating plurai pictures constituting a moving picture. In figure 43(a), part of plural pictures constituting one moving picture Mpt, i.e., pictures F(k)~F(k+2n-1) [k,n: integers], are shown. Display times t(k)~t(k+2n-1) are set on the respective pictures F(k)~F(k+2n-1) . As shown in figure 43(a), the respective pictures are successively arranged from one having earlier display time on a display time axis X indicating display times Tdis of the respective pictures, and these pictures are grouped for every predetermined number (n) of pictures. Each of these picture groups is called a GOP (Group of Pictures) , and this is a minimum unit of random access to coded data of a moving picture. In the following description, a picture group is sometimes abbreviated as a GOP. For example, an (i)th picture group Gp(i) is constituted by pictures F(k)~F(k+n-1) . An (i+1)th picture group Gp(i+1) is constituted by pictures F(n+k)~F(k+2n-1) . Each picture is divided into plural slices each comprising plural macroblocks. For example, a macroblock is a rectangle area having 16 pixels in the vertical direction and 16 pixels in the horizontal direction. Further, as shown in figure 43(b), a picture F(k+1) is divided into plural slices SL1~SLm [m: natural number]. A slice SL2 is constituted by plural macroblocks MB1~ MBr [r: natural number] as shown in figure 43(c). Figure 44 is a diagram for explaining coded data of a moving picture, illustrating a structure of a stream obtained by coding the respective pictures constituting the moving picture. A stream Smp is coded data corresponding to one image sequence (e.g., one moving picture). The stream Smp is composed of an area (common information area) Cstr wherein bit streams corresponding to common information such as a header are arranged, and an area (GOP area) Dgop wherein bit streams corresponding to the respective GOPs are arranged. The common information area Cstr includes sync data Sstr and a header Hstr corresponding to the stream. The GOP area Dgop includes bit streams Bg(1)~Bg(i- 1), Bg(i), Bg(i+1)~Bg(I) corresponding to picture groups (GOP) Gp(1)~Gp(i-1), Gp(i), Gp(i+1)~Gp(I) [i,I: integers]. Each bit stream corresponding to each GOP is composed of an area (common information area) Cgop wherein bit streams corresponding to common information such as a header are arranged, and an area (picture area) Dpct wherein bit streams corresponding to the respective pictures are arranged. The common information area Cgop includes sync data Sgop and a header Hgop corresponding to the GOP. A picture area Dpct of the bit stream Bg(i) corresponding to the picture group G(i) includes bit streams Bf(k'), Bf(k"+1), Bf(k'+2), Bf(k'+3), ..., Bf(k'+s) corresponding to pictures F(k')/ F(k'+1), F(k'+2), F(k'+3), ..., F(k'+s) [k',s: integers]. The pictures F(k'), F(k'+1), F(k'+2), F(k'+3), ..., F(k'+s) are obtained by rearranging, in coding order, the pictures F(k)~F(k+n-1) arranged in order of display times. Each bit stream corresponding to each picture is composed of an area (common information area) Cpct wherein bit streams corresponding to common information such as a header are arranged, and an area (slice area) Dslc wherein bit streams corresponding to the respective slices are arranged. The common information area Cpct includes sync data Spct and a header Hpct corresponding to the picture. For example, when the picture F(k'+1) in the arrangement in order of coding times (coding order arrangement) is the picture F(k+1) in the arrangement in order of display times (display order arrangement), the slice area Dslc in the bit stream Bf(k'+1) corresponding to the picture F(k'+1) includes bit streams Bs1~Bsm corresponding to the respective slices SL1~SLm. Each bit stream corresponding to each slice is composed of an area (common information area) Cslc wherein bit streams corresponding to common information such as a header are arranged, and an area (macroblock area) Dmb wherein bit streams corresponding to the respective macroblocks are arranged. The common information area Cslc includes sync data Sslc and a header Hslc corresponding to the slice. For example, when the picture F(k'+1) in the coding order arrangement is the picture F(k+1) in the display order arrangement, the macroblock area Dmb in the bit stream Bs2 corresponding to the slice SL2 includes bit streams Bm1~Bmr corresponding to the respective macroblocks MB1~MBr. As described above, coded data corresponding to one aoving picture (i.e., one image sequence) has a hierarchical structure comprising a stream layer corresponding to a stream Smp as the coded data, GOP layers corresponding to GOPs constituting the stream, picture layers corresponding to pictures constituting each of the GOPs, and slice layers corresponding to slices constituting each of the pictures. By the way, in moving picture coding methods such as MPEG (Moving Picture Experts Group)-1, MPEG-2, MPEG-4, ITU-T recommendation H.263, H.26L, and the like, a picture to be subjected to intra-picture coding is called an I picture, and a picture to be subjected to inter-picture predictive coding is called a P picture or a B picture. Hereinafter, definitions of an I picture, a P picture, and a B picture will be described. An I picture is a picture to be coded without referring to another picture. A P picture or B picture is a picture to be coded with reference to another picture. To be exact, a P picture is a picture for which either I mode coding or P mode coding can be selected when coding each block in the picture. A B picture is a picture for which one of I mode coding, P mode coding, and B mode coding can be selected when coding each block in the picture. The I mode coding is a process of performing intra-picture coding for a target block in a target picture without referring to another picture. The P mode coding is a process of performing inter-picture predictive coding for a target block in a target picture with reference to an already-coded picture. The B mode coding is a process of performing inter-picture predictive coding for a target block in a target picture with reference to two already-coded pictures. A picture to be referred to during the P mode coding or B mode coding is an I picture or a P picture other than the target picture, and it may be either a forward picture positioned forward the target picture or a backward picture positioned backward the target picture. However, there are three ways of combining two pictures to be referred to during the B mode coding. That is, there are three cases of B mode coding as follows: a case where two forward pictures are referred to, a case where two backward pictures are referred to, and a case where one forward picture and one backward picture are referred to. Figure 45 is a diagram for explaining a moving picture coding method such as MPEG described above. Figure 45 illustrates relationships between target pictures and the corresponding reference pictures (pictures to be referred to when coding the respective target pictures). Coding of the respective pictures F(k)~F(k+7), ..., F(k+17) ~F(k+21) constituting the moving picture is carried out with reference to other pictures as shown by arrows Z. To be specific, a picture at the end of one arrow Z is coded by inter-picture predictive coding with reference to a picture at the beginning of the same arrow Z. In figure 45, the pictures F(k)~F(k+7), —, F(k+17)~F(k+21) are identical to the pictures F(k)~F(k+4), ..., F(k+n-2)~F(k+n+4) , ..., F(k+2n-2), F(k+2n-l) shown in figure 43(a). These pictures are successively arranged from one having earlier display time on the display time axis X. The display times of the pictures F(k)~F(k+7), ..., F(k+17) ~F{k+21) are times t(k)~t(k+7), ..., t(k+17)~t(k+21). The picture types of the pictures F(k)~F(k+7) are I, B, B, P, B, B, P, B, and the picture types of the pictures F(k+17) ~F(k+21) are B, P, B, B, P. For example, when performing B mode coding for the second B picture F(k+1) shown in figure 45, the first I picture F(k) and the fourth P picture F(k+3) are referred to. Further, when performing P mode coding for the fourth P picture F(k+3) shown in figure 45, the first I picture F(k) is referred to. Although a forward picture is referred to in P mode coding of a P picture in figure 45, a backward picture may be referred to. Further, although a forward picture and a backward picture are referred to in B mode coding of a B picture in figure 45, two forward pictures or two backward pictures may be referred to. Furthermore, in a moving picture coding method such as MPEG- 4 or H.26L, a coding mode called "direct mode" may be selected when coding a B picture. Figures 46(a) and 46(b) are diagrams for explaining inter- picture predictive coding to be performed with the direct mode. Figure 46(a) shows motion vectors to be used in the direct mode. In figure 46(a), pictures P1, B2, B3, and P4 correspond to the pictures F(k+3)~F(k+6) [k=-2] shown in figure 45, and times t(l), t(2), t(3), and t(4) (t(1) Since the bit stream analysis unit 201, the mode decoding unit 203, and the prediction error decoding unit 202 operate in the same way as described for decoding of the picture P13, repeated description is not necessary. The motion compensation decoding unit 205 generates motion compensation data from the inputted information such as the motion vector. The bit stream analysis unit 201 outputs the motion vector and the reference picture index to the motion compensation decoding unit 205. The picture P11 is obtained by predictive coding using the pictures P7, B9 and P10 as candidate pictures for forward reference, and the picture P13 as a candidate picture for backward reference. At decoding the target picture, these reference candidate pictures have already been decoded, and are stored in the reference picture memory 207. Hereinafter, a description will be given of how the pictures stored in the reference picture memory 207 change with time, and a method for determining a reference picture, with reference to figure 3. The reference picture memory 207 is controlled by the memory control unit 204 on the basis of information Ih indicating what kind of reference has been carried out in coding P pictures and B pictures, which information is extracted from the header information of the bit stream. When decoding of the picture P11 is started, pictures P13, P4, P7, P10, and B9 are stored in the reference picture memory 207 as shown in figure 3. The picture B11 is decoded using the pictures P7, B9, and P10 as candidate pictures for forward reference, and the picture P13 as a backward reference picture. The decoded picture B11 is stored in the memory area where the picture P4 had been stored, because the picture P4 is not used as a candidate for a reference picture when decoding the picture B11 and the following pictures. In this case, which candidate picture has been referred to in detecting,the forward motion vector can be determined from the reference picture information added to the motion vector. To be specific, when the picture P10 has been referred to in coding the target block of the picture B11, information indicating that the candidate picture (picture P10) just previous to the target picture has been used as a reference picture (e.g., reference picture index [0]) is described in the bit stream of the target block. Further, when the picture B9 has been referred to in coding the target block, information indicating that the candidate picture which is two-pictures previous to the target picture has been used as a reference picture (e.g., reference picture index [1]) is described in the bit stream of the target block. Furthermore, when the picture P7 has been referred to in coding the target block of the picture P13, information indicating that the candidate picture which is three-pictures previous to the target picture has been used as a reference picture (e.g., reference picture index [2]) is described in, the bit stream of the target block. Accordingly, it is possible to know which one of the candidate pictures has been used as a reference picture in coding the target block, from the reference picture index. When the selected mode is bidirectional predictive coding, the motion compensation decoding unit 205 determines which one of the pictures P7, B9 and P10 has been used for forward reference, from the reference picture index. Then, the motion compensation decoding unit 205 obtains a forward motion compensation image from the reference picture memory 207 on the basis of the forward motion vector, and further, it obtains a backward motion compensation image from the reference picture memory 207 on the basis of the backward motion vector. Then, the motion compensation decoding unit 205 performs addition and averaging of the forward motion compensation image and the backward motion compensation image to generated a motion compensation image. Next, a process of generating a motion compensation image using forward and backward motion vectors will be described. (Bidirectional Prediction Mode) Figure 17 illustrates a case where the target picture to be decoded is the picture B11, and bidirectional predictive decoding is performed on a block (target block) BLa01 to be decoded, in the picture B11. Initially, a description will be given of a case where the forward reference picture is the picture P10, and the backward reference picture is the picture P13. In this case, the forward motion vector is a motion vector MVe01 indicating an area CRe01 in the picture P10, which area corresponds to the block BLa01. The backward motion vector is a motion vector MVg01 indicating an area CRg01 in the picture P13, which area corresponds to the block BLa01. Accordingly, the motion compensation decoding unit 205 obtains an image in the area CRe01 in the picture P10 as a forward reference image, and an image in the area CRgOl in the picture P13 as a backward reference image, from the reference picture memory 207, and performs addition and averaging of image data on the images in the both areas CRe01 and CRg01 to obtain a motion compensation image corresponding to the target block BLaOl. Next, a description will be given of a case where the forward reference picture is the picture B9, and the backward reference picture is the picture P13. In this case, the forward motion vector is a motion vector MVf01 indicating an area CRf01 in the picture B9, which area corresponds to the block BLa01. The backward motion vector is a motion vector MVgOl indicating an area CRg01 in the picture P13, which area corresponds to the block BLa01. Accordingly, the motion compensation decoding unit 205 obtains an image in the area CRf01 in the picture B9 as a forward reference image, and an image in the area CRg01 in the picture P13 as a backward reference image, from the reference picture memory 207, and performs addition and averaging of image data for the images in the both areas CRf01 and CRg01 to obtain a motion compensation image corresponding to the target block BLa01. (Direct Mode) Further, when the coding mode is the direct mode, the motion compensation decoding unit 205 obtains a motion vector (base motion vector) of a block that is included in the backward reference picture P13 for the target picture B11 and is placed relatively in the same position as the target block, which motion vector is stored in the motion vector storage unit 226. The motion compensation decoding unit 205 obtains a forward reference image and a backward reference image from the reference picture memory 207 by using the base motion vector. Then, the motion compensation decoding unit 205 performs addition and averaging of image data for the forward reference image and the backward reference image, thereby generating a motion compensation image corresponding to the target block. In the following description, a block in a picture, whose relative position with respect to a picture is equal to that of a specific block in another picture is simply referred to as a block which is located in the same position as a specific block in a picture. Figure 18(a) shows a case where the block BLa10 in the picture B11 is decoded in the direct mode with reference to the picture P10 that is just previous to the picture B11 (first example of direct mode decoding). A base motion vector to be used for direct mode decoding of the block BLa10 is a forward motion vector (base motion vector) MVh10 of a block (base block) BLg10 located in the same position as the block BLa10, which block BLg10 is included in the picture (base picture) P13 that is backward referred to when decoding the block BLalO. The forward motion vector MVh10 indicates an area CRh10 corresponding to the base block BLg10, in the picture P10 that is just previous to the picture B11. In this case, as a forward motion vector MVk10 of the target block BLa10 to be decoded, a motion vector which is parallel to the base motion vector MVhlO and indicates an area CRk10 included in the picture P10 and corresponding to the target block BLalO, is employed. Further, as a backward motion vector MVi10 of the target block BLa10 to be decoded, a motion vector which is parallel to the base motion vector MVhlO and indicates an area CRi10 included in the picture P13 and corresponding to the target block BLa10, is employed. Accordingly, the motion compensation decoding unit 205 obtains an image in the area CRk10 of the forward reference picture P10 as a forward reference image, and an image in the area CRi10 of the backward reference picture P13 as a backward reference image, from the reference picture memory 207, and performs addition and averaging of image data of the both images to obtain a motion compensation image (prediction image) corresponding to the target block BLa10. In this case, the magnitude (MVF) of the forward motion vector MVk10 and the magnitude (MVB) of the backward motion vector MVi10 are obtained by the above-described formulae (1) and (2), using the magnitude (MVR) of the base motion vector MVh10. The magnitudes MVF and MVB of the respective motion vectors show the horizontal component and vertical component of the motion vector, respectively. Further, TRD indicates a time-basis distance between the backward reference picture P13 for the target block BLa10 in the picture B11, and the picture P10 which is forward referred to when decoding the block (base block) BLg10 in the backward reference picture (base picture) P13. Further, TRF is the time- basis distance between the target picture B11 and the just- previous reference picture P10, and TRB is the time-basis distance between the target picture B11 and the picture P10 which is referred to when decoding the block BLg10 in the backward reference picture P13. Figure 18(b) shows a case where a block BLa20 in the picture B11 is decoded in the direct mode with reference to the picture P10 that is just previous to the picture B11 (second example of direct mode decoding). In this second example of direct mode decoding, in contrast with the first example of direct mode decoding shown in figure 18(a), a picture which is forward referred to in decoding the base block (i.e., a block placed in the same position as the target block, in the backward reference picture for the target block) is the picture P7. That is, a base motion vector to be used for direct mode decoding of the block BLa20 is a forward motion vector MVh20 of a block BLg20 located in the same position as the block BLa20, which block BLg20 is included in the picture P13 that is backward referred to when decoding the block BLa20. The forward motion vector MVh20 indicates an area CRh20 corresponding to the base block BLg20, in the picture P7 that is positioned forward the target picture B11. In this case, as a forward motion vector MVk20 of the target block BLa20 to be decoded, a motion vector, which is parallel to the base motion vector MVh20 and indicates an area CRk20 included in the picture P10 and corresponding to the target block BLa20, is employed. Further, as a backward motion vector MV120 of the target block BLa20 to be decoded, a motion vector, which is parallel to the base motion vector MVh20 and indicates an area CRi20 included in the picture P13 and corresponding to the target block BLa20, is employed. Accordingly, the motion compensation decoding unit 205 obtains an image in the area CRk20 of the forward reference picture P10 as a forward reference image, and an image in the area CRi20 of the backward reference picture P13 as a backward reference image, from the reference picture memory 207, and performs addition and averaging of image data of the both images to obtain a motion compensation image (prediction image) corresponding to the target block BLa20. In this case, the magnitude (MVF) of the forward motion vector MVk20 and the magnitude (MVB) of the backward motion vector MVi20 are obtained by the above-described formulae (1) and (2), using the magnitude (MVR) of the base motion vector MVh20, as described for the first example of direct mode decoding. Figure 19(a) shows a case where a block BLa30 in the picture B11 is decoded in the direct mode with reference to the picture P7 which is positioned forward the picture P10 that is positioned just previous to the picture B11 (third example of direct mode decoding). In this third example of direct mode decoding, in contrast with the first and second examples of direct mode coding shown in figures 18(a) and 18(b), a picture to be forward referred to in decoding the target block is not a picture just previous to the target picture, but a picture that is forward referred to in decoding the base block (a block in the same position as the target block) in the base picture. The base picture is a picture that is backward referred to in decoding the target block. That is, a base motion vector to be used in direct mode decoding of the block BLa30 is a forward motion vector MVh30 of a block BLg30 located in the same position as the block BLa30, which block BLg30 is included in the picture P13 that is backward referred to in decoding the block BLa30. The forward motion vector MVh30 indicates an area CRh30 corresponding to the base block BLg30, in the picture P7 that is positioned forward the target picture B11. In this case, as a forward motion vector MVk30 of the target block BLa30 to be decoded, a motion vector, which is parallel to the base motion vector MVh30 and indicates an area CRk30 included in the picture P7 and corresponding to the target block BLa30, is employed. Further, as a backward motion vector MVi30 of the target block BLa30 to be decoded, a motion vector, which is parallel to the base motion vector MVh30 and indicates an area CRi30 included in the picture P13 and corresponding to the target block BLa30, is employed. Accordingly, the motion compensation decoding unit 205 obtains an image in the area CRk30 of the forward reference picture P7 as a forward reference image, and an image in the area CRi30 of the backward reference picture P13 as a backward reference image, from the reference picture memory 207, and performs addition and averaging of image data of the both images to obtain a motion compensation image (prediction image) corresponding to the target block BLa30. In this case, the magnitude (MVF) of the forward motion vector MVk30 and the magnitude (MVB) of the backward motion vector MVi30 are obtained by the above-described formulae (2) and (3), using the magnitude (MVR) of the base motion vector MVh30. When the picture to be referred to in decoding the block BLg30 has already been deleted from the reference picture memory 207, the forward reference picture P10 that is timewise closest to the target picture is used as a forward reference picture in the third example of direct mode decoding. In this case, the third example of direct mode decoding is identical to the first example of direct mode decoding. Figure 19(b) shows a case where a block BLa40 in the picture B11 is decoded in the direct mode by using a motion vector whose magnitude is zero (fourth example of direct mode decoding). In this fourth example of direct mode decoding, the magnitude of the reference motion vector employed in the first and second examples shown in figures 18(a) and 18(b) is zero. In this case, as a forward motion vector MVk40 and a backward motion vector MVi40 of the block BLa40 to be decoded, a motion vector whose magnitude is zero is employed. That is, the forward motion vector MVk40 indicates an area (block) CRk40 of the same size as the target block, which area is included in the picture P10 and placed at the same position as the target block BLa40. Further, the backward motion vector MVi40 indicates an area (block) CRi40 of the same size as the target block, which area is included in the picture P13 and placed at the same position as the target block BLa40. Accordingly, the motion compensation decoding unit 205 obtains an image in the area (block) CRk40 of the forward reference picture P10 as a forward reference image, and an image in the area (block) CRi40 of the backward reference picture P13 as a backward reference image, from the reference picture memory 207, and performs addition and averaging of image data of the both images to obtain a motion compensation image (prediction image) corresponding to the target block BLa40. This method is applicable to, for example, a case where a block which is included in the picture P13 as a backward reference picture of the picture B11 and is located in the same position as the block BLa40 is a block having no motion vector like an intra-frame- coded block. The data of the motion compensation image thus generated is output to the addition unit 208. The addition unit 208 adds the inputted prediction error data and the motion compensation image data to generate decoded image data. The decoded image data so generated is output through the switch 210 to the reference picture memory 207, and the decoded image is stored in the reference picture memory 207. The memory control unit 204 controls the reference picture memory 207 on the basis of the header information Ih indicating what kind of reference has been carried out in coding the P pictures and B pictures extracted from the header information of the bit stream. As described above, the blocks in the picture B11 are successively decoded. When all of the blocks in the picture B11 have been decoded, decoding of the picture B12 takes place. In the B picture decoding described above, a specific block is sometimes treated as a skip block. Hereinafter, decoding of a skip block will be briefly described. When it is found that a specific block is treated as a skip block during decoding of an inputted bit stream, from a skip identifier or a block number information that is described in the bit stream, motion compensation, i.e., acquisition of a prediction image corresponding to a target block, is carried out in the direct mode. For example, as shown in figure 6(b), when the blocks MB(r+l) and MB(r+2) between the block MB(r) and the block MB(r+3) in the picture B11 are treated as skip blocks, the bit stream analysis unit 201 detects the skip identifier Sf from the bit stream Bs. When the skip identifier Sf is input to the mode decoding unit 223, the mode decoding unit 223 instructs the motion compensation decoding unit 205 to perform motion compensation in the direct mode. Then, the motion compensation decoding unit 205 obtains the prediction images of the blocks MB(r+1) and MB(r+2), on the basis of an image (forward reference image) of a block which is included in the forward reference picture P10 and placed in the same position as the block treated as a skip block, and an image (backward reference image) of a block in the same position as the block treated as a skip block, and then outputs the data of the prediction images to the addition unit 208. The prediction error decoding unit 202 outputs data whose value is zero, as difference data of the blocks treated as skip blocks. In the addition unit 208> since the difference data of the blocks treated as skip blocks is zero, the data of the prediction images of the blocks MB(r+1) and MB(r+2) are output to the reference picture memory 207 as decoded images of the blocks MB(r+1) and MB(r+2). Furthermore, in the direct mode processing shown in figure 18(a) (first example), the direct mode processing shown in figure 18(b) (second example), and the direction mode processing shown in figure 19(a) (third example), all of blocks whose difference data become zero are not necessarily treated as skip blocks. That is, a target block is subjected to bidirectional prediction using a picture that is positioned just previous to the target picture as a forward reference picture, and a motion vector whose magnitude is zero, and only when the difference data of the target block becomes zero, this target block may be treated as a skip block. In this case, when it is found, from the skip identifier or the like in the bit stream Bs, that a specific block is treated as a skip block, motion compensation should be carried out by bidirectional prediction whose motion is zero, using a just- previous reference picture as a forward reference picture. (Decoding Process for Picture B12) Since the bit stream analysis unit 201, the mode decoding unit 223, and the prediction error decoding unit 202 operate in the same way as described for decoding of the picture P10, repeated description is not necessary. The motion compensation decoding unit 205 generates motion compensation image data from the inputted information such as the motion vector. The motion vector MV and the reference picture index Rp are input to the motion compensation decoding unit 205. The picture P12 has been coded using the pictures P7, P10 and B11 as candidate pictures for forward reference, and the picture P13 as a candidate picture for backward reference. At decoding the target picture, these candidate pictures have already been decoded, and are stored in the reference picture memory 207. The timewise change of the pictures stored in the reference picture memory 207, and the method for determining a reference picture are identical to those in the case of decoding the picture B11 described with respect to figure 3. When the coding mode is bidirectional predictive coding, the motion compensation decoding unit 205 determines which one of the pictures P7, P10 and B11 has been used for forward reference, from the reference picture index. Then, the motion compensation decoding unit 205 obtains a forward reference image from the reference picture memory 207 on the basis of the forward motion vector, and further, it obtains a backward reference image from the reference picture memory 207 on the basis of the backward motion vector. Then, the motion compensation decoding unit 205 performs addition and averaging of image data of the forward reference image and the backward reference image to generated a motion compensation image corresponding to the target block. (Bidirectional Prediction Mode) Figure 20 illustrates a case where the target picture to be decoded is the picture B12, and bidirectional predictive decoding is performed for a block (target block) BLa02 to be decoded, in the picture B12. Initially, a description will be given of a case where the forward reference picture is the picture B11, and the backward reference picture is the picture P13. In this case, the forward motion vector is a motion vector MVe02 indicating an area CRe02 in the picture B11 corresponding to the block BLa02. The backward motion vector is a motion vector MVg02 indicating an area CRg02 in the picture P13 corresponding to the block BLa02. Accordingly, the motion compensation decoding unit 205 obtains an image in the area CRe02 in the picture B11 as a forward reference image, and an image in the area CRg02 in the picture P13 as a backward reference image, from the reference picture memory 207, and performs addition and averaging of image data of the images in the both areas CRe02 and CRg02 to obtain a motion compensation image corresponding to the target block BLa02. Next, a description will be given of a case where the forward reference picture is the picture P10, and the backward reference picture is the picture P13. In this case, the forward motion vector is a motion vector MVf02 indicating an area CRf02 in the picture P10, corresponding to the block BLa02. The backward motion vector is a motion vector MVg02 indicating an area CRg02 in the picture P13, corresponding to the block BLa02. Accordingly, the motion compensation decoding unit 205 obtains an image in the area CRf02 in the picture P10 as a forward reference image and an image in the area CRg02 in the picture P13 as a backward reference image from the reference picture memory 207, and performs addition and averaging of image data of the images in the both areas CRf02 and CRg02 to obtain a motion compensation image corresponding to the target block BLa02. (Direct Mode) Further, when the coding mode is the direct mode, the motion compensation decoding unit 205 obtains a motion vector (base motion vector) of a reference block (a block whose relative position is the same as that of the target position) in the backward reference picture P13 for the target picture B12, which motion vector is stored in the motion vector storage unit 226. The motion compensation decoding unit 205 obtains a forward reference image and a backward reference image from the reference picture memory 207 by using the base motion vector. Then, the motion compensation decoding unit 205 performs addition and averaging of image data of the forward reference image and the backward reference image, thereby generating amotion compensation image corresponding to the target block. Figure 21(a) shows a case where the block BLa50 in the picture B12 is decoded in the direct mode with reference to the picture B11 that is just previous to the picture B12 (first example of direct mode decoding). A base motion vector to be used for direct mode decoding of the block BLa50 is a forward motion vector MVj50 of the base block (i.e., the block BLg50 placed in the same position as the block BLa50) in the picture P13 that is backward referred to when decoding the block BLa50. The forward motion vector MVj50 indicates an area CRj50 corresponding to the base block BLg50 in the picture P10 that is positioned forward and close to the picture B11. In this case, as a forward motion vector MVk50 of the target block BLa50 to be decoded, a motion vector which is parallel to the base motion vector MVj50 and indicates an area CRk50 included in the picture B11 and corresponding to the target block BLa50, is employed. Further, as a backward motion vector MVi50 of the target block BLa50 to be decoded, a motion vector which is parallel to the base motion vector MVj50 and indicates an area CRi50 included in the picture P13 and corresponding to the target block BLa50, is employed. Accordingly, the motion compensation decoding unit 205 obtains an image in the area CRk50 of the forward reference picture B11 as a forward reference image and an image in the area CRi50 of the backward reference picture P13 as a backward reference image from the reference picture memory 207, and performs addition and averaging of image data of the both images 152 to obtain a motion compensation image (prediction image) corresponding to the target block BLa50. In this case, the magnitude (MVF) of the forward motion vector MVkSO and the magnitude (MVB) of the backward motion vector MVi50 are obtained by the above-described formulae (1) and (2) using the magnitude (MVR) of the base motion vector MVh10. The magnitudes MVF and MVB of the respective motion vectors show the horizontal component and vertical component of the motion vector, respectively. Figure 21(b) shows a case where a block BLa60 in the picture B12 is decoded in the direct mode with reference to the picture B11 that is positioned forward the picture B12 (second example of direct mode decoding). In this second example of direct mode decoding, in contrast with the first example of direct mode decoding shown in figure 21(a), a picture which is forward referred to in decoding the base block (i.e., a block placed in the same position as the target block, in the backward reference picture for the target block) is the picture P7. That is, a base motion vector to be used for direct mode decoding of the block BLa60 is a forward motion vector MVj60 of the reference block (the block BLg60 in the same position as the block BLa60) in the picture P13 that is backward referred to when decoding the block BLa60. The forward motion vector MVj60 indicates an area CRj60 corresponding to the base block BLg60, in the picture P7 that is positioned forward the target picture B12. In this case, as a forward motion vector MVk60 of the target block BLa60 to be decoded, a motion vector, which is parallel to the base motion vector MVj60 and indicates an area CRk60 included in the picture B11 and corresponding to the target block BLa60, is employed. Further, as a backward motion vector MVi60 of the target block BLa60 to be decoded, a motion vector, which is parallel to the base motion vector MVj60 and indicates an area CRi60 included in the picture P13 and corresponding to the target block BLa60, is employed. Accordingly, the motion compensation decoding unit 205 obtains an image in the area CRk60 of the forward reference picture B11 as a forward reference image and an image in the area CRi60 of the backward reference picture P13 as a backward reference image from the reference picture memory 207, and performs addition and averaging of image data of the both images to obtain a motion compensation image (prediction image) corresponding to the target block BLa60. In this case, the magnitude (MVF) of the forward motion vector MVk60 and the magnitude (MVB) of the backward motion vector MVi60 are obtained by the above-described formulae (1) and (2), using the magnitude (MVR) of the base motion vector MVj60, as described for the first example of direct mode decoding. Figure 22(a) shows a case where a block BLa70 in the picture B12 is decoded in the direct mode with reference to the picture P7 which is positioned forward the forward picture P10 that is closest to the picture B12 (third example of direct mode decoding). In this third example of direct mode decoding, in contrast with the first and second examples of direct mode coding shown in figures 21(a) and 21(b), a picture to be forward referred to in decoding the target block is not a picture just previous to the target picture, but a picture that is forward referred to in decoding the base block in the base picture. The base picture is a picture that is backward referred to in decoding the target block, That is, a base motion vector to be used in direct mode decoding of the block BLa70 is a forward motion vector MVj70 of a base block BLg70 (a block in the same position as the block BLa70) in the picture P13 that is backward referred to in decoding the block BLa70. The forward motion vector MVj70 indicates an area CRj70 corresponding to the base block BLg70 in the picture P7 that is positioned forward the target picture B12. In this case, as a forward motion vector MVk70 of the target block BLa70 to be decoded, a motion vector which is parallel to the base motion vector MVj70 and indicates an area CRk70 included in the picture P7 and corresponding to the target block BLa70, is employed. Further, as a backward motion vector MVi70 of the target block BLa70, a motion vector which is parallel to the base motion vector MVj70 and indicates an area CRi70 included in the picture P13 and corresponding to the target block BLa70, is employed. Accordingly, the motion compensation decoding unit 205 obtains an image in the area CRk70 of the forward reference picture P7 as a forward reference image and an image in the area CRi70 of the backward reference picture P13 as a backward reference image from the reference picture memory 207, and performs addition and averaging of image data of the both images to obtain a motion compensation image (prediction image) corresponding to the target block BLa70. In this case, the magnitude (MVF) of the forward motion vector MVk70 and the magnitude (MVB) of the backward motion vector MVi70 are obtained by the above-described formulae (2) and (3), using the magnitude (MVR) of the base motion vector MVj70. When the picture to be referred to in decoding the block BLg70 has already been deleted from, the reference picture memory 207, the forward reference picture P10 that is timewise closest to the target picture is used as a forward reference picture in the third example of direct mode decoding. In this case, the third example of direct mode decoding is identical to the first example of direct mode decoding. Figure 22(b) shows a case where a block BLa80 in the picture B12 is decoded in the direct mode by using a motion vector whose magnitude is zero (fourth example of direct mode decoding). In this fourth example of direct mode decoding, the magnitude of the reference motion vector employed in the first and second examples shown in figures 21(a) and 21(b) is zero. In this case, as a forward motion vector MVk80 and a backward motion vector MVi80 of the block BLa80 to be decoded, a motion vector whose magnitude is zero is employed. That is, the forward motion vector MVk80 indicates an area (block) CRk80 of the same size as the target block, which area is included in the picture B11 and placed at the same position as the target block BLa80. Further, the backward motion vector MVi80 indicates an area (block) CRi80 of the same size as the target block, which area is included in the picture P13 and placed at the same position as the target block BLa80. Accordingly, the motion compensation decoding unit 205 obtains an image in the area (block) CRk80 of the forward reference picture B11 as a forward reference image and an image in the area (block) CRi80 of the backward reference picture P13 as a backward reference image from the reference picture memory 207, and performs addition and averaging of image data of the both images to obtain a motion compensation image (prediction image) corresponding to the target block BLa80. This method is applicable to, for example, a case where a block which is included in the picture P13 as a backward reference picture of the picture B11 and is located in the same position as the block BLa80 is a block having no motion vector like an intra-frame- coded block. The data of the motion compensation image thus generated is output to the addition unit 208. The addition unit 208 adds the inputted prediction error data and the motion compensation image data to generate decoded image data. The decoded image data so generated is output through the switch 210 to the reference picture memory 207. As described above, the blocks in the picture B12 are successively decoded. The image data of the respective pictures stored in the reference picture memory 207 are rearranged in order of time to be output as output image data Od. Thereafter, the pictures following the picture B12, which are arranged in decoding order times as shown in figure 16(a), are successively decoded according to the picture type, in like manner as described for the pictures P13, B11, and B12. Figure 16(b) shows the pictures rearranged in order of display times. During decoding of the inputted bit stream, if it is found that a specific block is treated as a skip block, from a skip identifier or a block number information that is described in the bit stream, motion compensation, i.e., acquisition of a prediction image corresponding to a target block, is carried out in the direct mode as in the case of decoding the picture B11. As described above, in the moving picture decoding apparatus 20 according to the second embodiment, when decoding a block in a B picture, a prediction image corresponding to the target block is generated using an already-decoded P picture and an already- decoded B picture as candidate pictures for forward reference, on the basis of information (reference picture index) indicating candidate pictures which are forward referred to in coding the target block, which information is included in the bit stream corresponding to the target block to be decoded. Therefore, it is possible to correctly decode a block in a target B picture which has been coded using a B picture as a candidate picture for forward reference. Further, in the moving picture decoding apparatus 20, when a target block to be decoded which is included in a B picture has been coded in the direct mode, a motion vector of the target block is calculated on the basis of a motion vector of a block that is placed in the same position as the target block. Therefore, it is not necessary for the decoding end to obtain the information indicating the motion vector of the block coded in the direct mode, from the coding end. Furthermore, in the moving picture decoding apparatus 20, the data of the already-decoded pictures which are stored in the reference picture memory are managed on the basis of the information indicating the candidate pictures which are used in coding p pictures and B pictures, which information is included as header information in the bit stream. For example, at the completion of decoding one picture, data of pictures which are not to be Used as reference pictures in decoding the following pictures are successively deleted, whereby the picture memory can be used with efficiency. Further, when decoding a target block in a P picture, it is possible to determine which one of plural candidate pictures is used as a reference picture (i.e., which one of the candidate pictures is referred to in detecting the motion vector of the target block to be decoded) from the reference picture information added to the motion vector information. Likewise, when decoding a target block in a B picture, it is possible to determine which one of plural candidate pictures for forward reference is used as a reference picture (i.e., which one of the candidate pictures is referred to in detecting the forward motion vector of the target block to be decoded) from the reference picture information added to the motion vector information. While in this second embodiment the direct mode is used as one of the plural coding modes for B pictures, the direct mode is not necessarily used as the coding mode for B pictures. In this case, the motion vector storage unit 226 in the moving picture decoding apparatus 20 is dispensed with. Further, while in this second embodiment four specific methods are described as examples of direct mode (i.e., the first example shown in figure 18(a) or 21(a), the second example shown in figure 18(b) or 21(b), the third example shown in figure 19 (a) or 22(a), and the fourth example shown in figure 19(b) or 22(b)), the decoding apparatus performs decoding using a method suited to a coding method which is used as direct mode by the coding apparatus. More specifically, when plural methods are employed as direct mode, the decoding apparatus performs decoding, using information indicating which one of the plural methods is used as specific direct mode, that is described in the bit stream. In this case, the operation of the motion compensation decoding unit 205 varies according to the information. For example, when this information is added in block units for motion compensation, the mode decoding unit 223 determines which one of the four methods mentioned above is used as direct mode in coding, and notifies the motion compensation decoding unit 205 of the determined method. The motion compensation decoding unit 2 05 performs appropriate motion compensation predictive decoding according to the determined method of direct mode. Further, when the information (DM mode information) indicating which one of the plural methods is used as direct mode is described in the header of the entire sequence, the GOP header, the picture header, or the slice header, the DM mode information is transferred for every sequence, GOP, picture, or slice, from the bit stream analysis unit 201 to the motion compensation decoding unit 205, and the motion compensation decoding unit 205 changes the operation. While in this second embodiment two B pictures are placed between an I picture and a P picture or between adjacent P pictures, the number of continuous B pictures may be three or four. Further, while in this second embodiment three pictures are used as candidate pictures for a forward reference picture for a P picture, the number of reference candidate pictures for a P picture may be other than three. Furthermore, while in this second embodiment two I or P pictures and one B picture are used as candidate pictures for a forward reference picture in decoding a B picture, forward reference candidate pictures in decoding a B picture are not restricted thereto. Moreover, in this second embodiment, as a method for managing the reference picture memory in decoding the picture P13, picture B11, and picture B12, a method of collectively managing the P pictures and B pictures to be used as candidates of a reference picture, as shown in figure 3, is described. However, the reference picture memory managing method may be any of the four methods which are described for the first embodiment with reference to figures 11 to 14, wherein all of the pictures to be used as candidates for a reference picture are separated into P pictures and B pictures to be managed. In this case, the reference picture memory 207 has memory areas for six pictures, i.e., P picture memory areas (#1)~(#4), and B picture memory areas (#1) and (#2) . Further, these six memory areas are not necessarily formed in one reference picture memory, but each of the six memory areas may be constituted by one independent reference picture memory. Further, when the coding end employs a reference picture index assigning method wherein it is determined, for each picture to be coded, which of the P picture memory area and the B picture memory area is given priority in assigning reference picture indices as shown in figure 14, the moving picture decoding apparatus can easily identify a picture which is used as a reference picture among plural candidate pictures, on the basis of the reference picture indices, by using information described in the bit stream, which indicates the memory area taking priority. For example, when the target picture to be decoded is the picture B11, since the forward reference picture that is timewise closest to the target picture is the picture P10, reference picture indices are assigned to the pictures stored in the P picture memory with priority. Accordingly, a reference picture index [0] is added as header information to the bit stream of the target block when the picture P10 is used as a reference picture in coding the target block of the picture B11. Likewise, a reference picture index [1] is added as header information when the picture P7 is used as a reference picture, and a reference picture index [2] is added as header information when the picture B9 is used as a reference picture. Accordingly, the moving picture decoding apparatus can know which candidate picture is used as a reference picture in coding the target block, according to the reference picture index. In this case, since information indicating that reference picture indices are assigned to the candidate pictures in the P picture memory with priority is included as header information in the bit stream, identification of the reference picture is further facilitated by using this information. Further, when the target picture to be decoded is the picture B12, since the forward reference picture that is timewise closest to the target picture is the picture B11, reference picture indices are assigned to the pictures stored in the B picture memory with priority. Accordingly, a reference picture index [0} is added as header information to the bit stream of the target block when the picture B11 is used as a reference picture in coding the target block of the picture B12. Likewise, a reference picture index [1] is added as header information when the picture P10 is used as a reference picture, and a reference picture index [2] is added as header information when the picture P7 is used as a reference picture. Accordingly, the moving picture decoding apparatus can know which candidate picture is used as a reference picture in coding the target block, according to the reference picture index. In this case, since information indicating that reference picture indices are assigned to the candidate pictures in the B picture memory with priority is included as header information in the bit stream, identification of the reference picture is further facilitated by using this information. Furthermore, there are cases where, at the coding end, one of the above-mentioned five methods for managing the reference picture memory (refer to figures 3, 11 to 14) is previously selected, or some of these five methods are used by switching them. For example, when the coding end employs some of the plural methods by switching them, the moving picture decoding apparatus can determine the reference picture index, according to information indicating which method is used for each picture, that is described in the bit stream. Furthermore, in this second embodiment, the five methods for managing the reference picture memory (refer to figures 3, 11 to 14) are described for the case where there are three reference candidate pictures for a P picture, and there are two P pictures and one B picture as forward reference candidate pictures for a B picture. However, the five methods for managing the reference picture memory are also applicable to cases where the number of reference candidate pictures is different to those mentioned above. When the number of reference candidate pictures is different from those mentioned for the second embodiment, the capacity of the reference picture memory is also different from that described for the second embodiment. Moreover, in this second embodiment, in the method of managing the reference picture memory wherein the stored reference candidates are separated into P pictures and B pictures (four examples shown in figures 11 to 14), the P pictures are stored in the P picture memory area while the B pictures are stored in the B picture memory area. However, a short-term picture memory and a long-term picture memory which are defined in H.263++ may be used as memory areas where pictures are stored. For example, the short-term picture memory and the long-term picture memory may be used as a P picture memory area and a B picture memory area, respectively. [Embodiment 3] Figure 23 is a block diagram illustrating a moving picture coding apparatus 30 according to a third embodiment of the present invention. The moving picture coding apparatus 30 can switch, according to a control signal supplied from the outside, a method for assigning reference picture indices to candidate pictures, between a method of assigning reference picture indices to candidate pictures according to an initialized rule (default assignment method), and an adaptive assignment method of assigning reference picture indices to candidate pictures by the default assignment method and, further, adaptively changing the assigned reference picture indices according to the coding status. To be specific, one operation mode of the moving picture coding apparatus 30 according to the third embodiment is the operation of the moving picture coding apparatus 10 according to the first embodiment. In other words, when the default assignment method is selected as a reference picture index assignment method of the moving picture coding apparatus 30, the moving picture coding apparatus 30 performs the same processing as that of the moving picture coding apparatus 10. Hereinafter, the moving picture coding apparatus 30 will be described in detail. The moving picture coding apparatus 30 is provided with a coding control unit 130, instead of the coding control unit 110 of the moving picture coding apparatus 10 according to the first embodiment. The coding control unit 130 switches, according to an external control signal Cont, a method for assigning reference picture indices to candidate pictures, between a method of assigning reference picture indices according to an initialized rule (default assignment method), and a method including a first step of assigning reference picture indices to candidate pictures by the default assignment method, and a second step of adaptively changing the reference picture indices which are assigned to the candidate pictures by the default assignment method (adaptive assignment method). Further, the coding control unit 130 includes a detection unit (not shown) which detects, for every target picture to be coded, coding efficiency in a case where each of plural reference candidate pictures is used as a reference picture. The coding control unit 130 changes the reference picture index which is assigned to each candidate picture by the default assignment method, according to the coding efficiency detected by the detection unit. More specifically, the coding control unit 130 changes the reference picture index which is assigned to each candidate picture by the default assignment method, such that, among plural candidate pictures for a target picture, a candidate picture which provides a higher coding efficiency of the target picture when it is used as a reference picture is given a smaller reference picture index. Then, the mode selection unit 139 selects, in the direct mode, a picture that is assigned a reference picture index [0], as a forward reference picture for a target block. In a predictive coding mode other than the direct mode such as the bidirectional predictive coding mode, the mode selection unit 139 selects a reference picture from among plural candidate pictures according to the coding efficiency. Other components of the moving picture coding apparatus 30 according to the third embodiment are identical to those of the moving picture coding apparatus 10 according to the first embodiment. Hereinafter, the operation will be described. In the moving picture coding apparatus 30, when the default assignment method is selected as a method for assigning reference picture indices to candidate pictures according to the external control signal Cont, the operation of the moving picture coding apparatus 30 is identical to the operation of the moving picture coding apparatus 10 according to the first embodiment. On the other hand, when the adaptive assignment method is selected as a method for assigning reference picture indices to candidate pictures according to the external control signal Cont, the moving picture coding apparatus 30 performs, in the first step, assignment of reference picture indices in like manner as described for the moving picture coding apparatus 10. When the adaptive assignment method is selected, the moving picture coding apparatus 30 performs, in the second step, adaptive change of the reference picture indices that are assigned by the default assignment method. Hereinafter, a description will be given of specific methods of assigning reference picture indices in the case where the adaptive assignment method is selected. In the following description, it is assumed that a target picture is the picture B12. Initially, in the first step, as shown in figure 3, reference picture indices are assigned to candidate pictures for forward reference such that a smaller reference picture index is assigned to a candidate picture that is closer to the target picture. That is, a reference picture index [1] is assigned to the reference picture P10, a reference picture index [0] is assigned to the reference picture B11, and a reference picture index [2] is assigned to the reference picture P7. Next, in the second step, as shown in figure 24, the reference picture index [1] of the reference picture P10 is changed to [0], and the reference picture index [0] of the reference picture B11 is changed to [1]. Such rewriting of reference picture indices is carried out for every target picture, according to the coding efficiency. Further, the moving picture coding apparatus 30 inserts information indicating which of the default assignment method and the adaptive assignment method is set as an assignment method, as header information, in the bit stream. Further, when the adaptive assignment method is set, information indicating how the assignment of reference picture indices is carried out is also inserted as header information in the bit stream. As described above, in this third embodiment, the reference picture index of the candidate picture which is to be used as a forward reference picture in the direction mode, can be changed to [0] . That is, since, in the first embodiment, a smaller reference picture index is given to a reference candidate picture that is timewise closer to the target picture, only the picture B11 that is timewise closest to the target picture B12 can be referred to in the direct mode. In this third embodiment, however, any picture other than the picture B11 closest to the target picture B12 can be used as a forward reference picture, if the coding efficiency is improved. Further, in this case, since the picture to be referred to in coding the picture B12 in the direct mode is not the picture B11 but the picture B10, decoding of the picture B11 becomes unnecessary. Accordingly, as shown in figure 25(a), a B picture immediately after a P picture can be processed without decoding it, whereby speedup of decoding is achieved when the picture B11 is not necessary. Further, since decoding can be carried out even when the data of the picture B11 is lost due to transmission error or the like, reliability of decoding is improved. As described above, when a reference picture index can be arbitrarily assigned to a candidate picture to intentionally determine a picture to be referred to in the direct mode, a predetermined picture can be processed without decoding it, as shown in figure 25(a). Furthermore, even when three B pictures are placed between P pictures as shown in figure 25(b), a predetermined picture can be processed without decoding it. Therefore, when a picture that is not needed by the user is previously known at the coding end, such picture can be omitted to reduce the processing time in decoding. In figure 25(b), even when the picture B3 is not decoded, other pictures can be decoded. That is, in the assignment method of the first embodiment, since the picture B4 refers to the picture B3 in the direct mode, the picture B3 must be decoded to decode the picture B4. In this third embodiment, however, since a picture to be referred to in the direct mode can be arbitrarily set, decoding of the picture B3 can be dispensed with. Furthermore, in this third embodiment, assignment of reference picture indices is carried out such that a smaller reference picture index is assigned to a candidate picture that is timewise closer to the target picture, and a reference picture to be used in the direct mode is determined according to the reference picture indices. Therefore, the coding efficiency can be improved by a reduction in the motion vector, and further, the processing time can be reduced. Furthermore, when the target block is processed in the direct mode at the decoding end, since the forward reference candidate picture to which the reference picture index [0] is assigned is immediately used as a reference picture, decoding time can be reduced. Furthermore, while in this third embodiment a candidate picture whose reference picture index should be changed to [0] is determined according to the coding efficiency, a reference picture index of a picture which is most likely to be referred to, e.g., a P picture that is timewise closest to the target picture, may be changed to [0]. Moreover, while in this third embodiment a picture to be referred to in the direct mode is a picture whose reference picture index is [0], the present invention is not restricted thereto. For example, information indicating that a picture is to be referred to in the direct mode is coded, and decoding may be carried out in the direct mode on the basis of this information. [Embodiment 4] Figure 26 is a block diagram for explaining a moving picture decoding apparatus 40 according to a fourth embodiment of the present invention. The moving picture decoding apparatus 40 receives the bit stream outputted from the moving picture coding apparatus 30 of the third embodiment, and performs decoding of each picture, on the basis of information indicating which of the default assignment method and the adaptive assignment method should be used when assigning reference picture indices (assignment method instruction information), which information is included in the bit streanu That is, one operation mode of the moving picture decoding apparatus 40 according to the fourth embodiment is the operation of the moving picture coding apparatus 20 according to the second embodiment. In other words, when the default assignment method is used as a reference picture index assignment method in the moving picture decoding apparatus 40, the operation of the moving picture decoding apparatus 40 is identical to that of the moving picture decoding apparatus 20. Hereinafter, the moving picture decoding apparatus 40 will be described in detail. The moving picture decoding apparatus 40 is provided with a memory control unit 244, instead of the memory control unit 204 of the moving picture decoding apparatus 20 according to the second embodiment. The memory control unit 244 performs memory management according to either the default assignment method or the adaptive assignment method, on the basis of the assignment method instruction information included in the bit stream as header information. Other components of the moving picture decoding apparatus 40 according to the fourth embodiment are identical to those of the moving picture decoding apparatus 20 according to the second embodiment. Hereinafter, the operation will be described. The moving picture decoding apparatus 40 operates in accordance with the assignment method instruction information that is included as header information in the bit stream supplied from the moving picture coding apparatus 30. That is, when the default assignment method is selected as a reference picture index assignment method at the coding end, i.e., when information indicating that the default assignment method is selected is included in the bit stream, the moving picture decoding apparatus 40 operates in the same manner as the moving picture decoding apparatus 20 of the second embodiment. On the other hand, when the adaptive assignment method is selected as a reference picture index assignment method at the coding end, i.e., when information indicating that the adaptive assignment method is selected is included in the bit stream, the moving picture decoding apparatus 40 operates in accordance with the adaptive assignment method. In this case, since information indicating how the assignment of reference picture indices is carried out is also included as header information in the bit stream, assignment of reference picture indices is carried out according to this information. Hereinafter, a description will be given of the operation of the moving picture decoding apparatus 40 in the case where the adaptive assignment method is selected. In the reference picture memory 207, as shown in figure 24, reference candidate pictures stored in the respective memory area are rewritten every time a target picture is processed. To be specific, when the target picture to be decoded is the picture B12, decoding of a target block in the picture B12 is carried out with reference to a reference picture that is selected from candidate pictures according to the header information of the target block. For example, when the coding mode for the target block is the bidirectional predictive mode, a candidate picture which is given the same reference picture index as the reference picture index that is included in the header information of the target block, is selected as a forward reference picture from among the candidate pictures P10, B11, and P7. When the reference picture index included in the header information of the target block is [1], the candidate picture B11 is selected as a forward reference picture. Then, the target block is subjected to bidirectional predictive decoding with reference to the candidate picture B11 as a forward reference picture, and the picture P13 as a backward reference picture. Further, when the decoding mode of the target block is the direct mode, a candidate picture (picture P10) which is given the reference picture index [0] is selected as a forward reference picture from among the candidate pictures P7, P10, and B9. Then, the target block is decoded with reference to the candidate picture P10 as a forward reference picture, and the picture P13 as a backward reference picture. As described above, according to the fourth embodiment, the reference picture memory 207 is managed as shown in figure 24, that is, memory management is carried out using, as the reference picture indices of the respective candidate pictures, those obtained by changing the reference picture indices assigned by the default assignment method, according to the coding status. Therefore, it is possible to realize a decoding method adaptive to a coding method in which the reference picture indices of the candidate pictures are rewritten according to the coding efficiency. That is, since, in the second embodiment, a smaller reference picture index is given to a reference candidate picture that is timewise closer to the target picture, only the picture B11 that is timewise closest to the target picture B12 can be used as a reference picture in the direct mode. In this fourth embodiment, however, a picture other than the picture B11 closest to the target picture B12 can be used as a forward reference picture. Further, in this case, since the picture to be referred to in decoding a block in the picture B12 in the direct mode is not the picture B11 but the picture B10, decoding of the picture B11 becomes unnecessary. Accordingly, as shown in figure 25(a), a B picture immediately after a P picture can be processed without decoding it, whereby speedup of decoding is achieved when the picture B11 is not necessary. Further, since decoding can be carried out even when the data of the picture B11 is lost due to transmission error or the like, reliability of decoding is improved. As described above, when a reference picture index to be assigned to each reference candidate picture is arbitrarily selected according to the coding status to intentionally determine a picture to be referred to in the direct mode, a predetermined picture can be processed without decoding it as shown in figure 25(a). Furthermore, even when three B pictures are placed between P pictures as shown in figure 25(b), a predetermined picture can be processed without decoding it. Therefore, if a picture that is not needed by the user is previously known at the coding end, such picture can be omitted to reduce the processing time for decoding. In figure 25(b), even when the picture B3 is not decoded, other pictures can be decoded. That is, since, in the second embodiment, the picture B4 is decoded with reference to the picture B3 in the direct mode, the picture B3 must be decoded. In this fourth embodiment, however, since a picture to be referred to in the direct mode is arbitrarily set at the coding end, decoding of the picture B3 can be dispensed with. Furthermore, when the target block is processed in the direct mode at the decoding end, since the forward reference candidate picture to which the reference picture index [0] is assigned is immediately used as a reference picture, decoding time can be reduced. While in the first to fourth embodiments a B picture is not referred to when coding or decoding a P picture, a B picture may be referred to when coding or decoding a P picture. Further, while in the first to fourth embodiments a time- basis distance between pictures is calculated according to the display times of the respective pictures, it may be calculated according to information other than time information such as the display times of pictures. For example, a counter value that is incremented every time a picture is processed is set, and a time-basis distance between pictures may be calculated according to this count value. To be specific, when time information is included in both of a video stream and an audio stream corresponding to a single contents, it is not easy to manage video data and audio data on the basis of the time information so as to maintain synchronization between these data, because a unit of time information is small. However, management considering synchronization between video data and audio data is facilitated by managing arrangement of the respective pictures with the counter value. Furthermore, in the first to fourth embodiments, a header section and a data section in a data processing unit, such as a GOP or a picture, are not separated from each other, and they are included in a bit stream corresponding to each data processing unit to be transferred. However, the header section and the data section may be separated from each other to be transferred in different streams. For example, when a stream is transferred in units of data transfer such as packets into which the stream is divided, a header section and a data section corresponding to a picture may be transferred separately from each other. In this case, the header section and the data section are not always included in the same stream. However, in data transfer using packets, even when the header section and the data section are not continuously transferred, the corresponding header section and data section are merely transferred in different packets, and the relationship between the corresponding header section and data section is stored in header information of each packet, and therefore, it is substantially identical to that the header section and the data section are included in the same bit stream. Furthermore, while in the first to fourth embodiments the reference picture indices are used as information for identifying which one of plural reference candidate pictures is referred to in coding a target block, the reference picture indices may be used as information indicating the positions of plural forward reference candidate pictures for a target picture to be coded or decoded. To be specific, in the reference picture index assignment methods according to the first and second embodiments or the default assignment methods according to the third and fourth embodiments, reference picture indices are assigned to the plural forward reference candidate pictures such that a smaller reference picture index is assigned to a candidate picture closer to the target picture, and therefore, the position of each forward reference candidate picture (i.e., the ordinal rank of each forward reference candidate picture in nearness to the target picture, among all forward reference candidate pictures) can be detected according to the reference picture index assigned to the forward reference candidate picture. Furthermore, position identification information indicating the positions of the respective pictures constituting a moving picture on the display time axis may be included in the bit stream corresponding to the moving picture, separately from the reference picture indices indicating the relative positions of the forward reference candidate pictures. The position identification information is different from the time information indicating the display times of pictures, and it is information specifying the relative positions of the respective pictures. Moreover, in the first to fourth embodiments, a picture that is to be backward referred to when coding a block in a target picture to be coded or decoded (backward reference picture for a target picture) is used as a base picture in the direct mode. However, a base picture to be used in the direct mode may be an already-processed picture other than the backward reference picture for the target picture, e.g., a picture to be forward referred to when coding the block in the target picture. [Embodiment 5] Figure 27 is a block diagram for explaining a moving picture coding apparatus 50 according to a fifth embodiment of the present invention. The moving picture coding apparatus 50 according to the fifth embodiment is different from the moving picture coding apparatus 10 according to the first embodiment in candidate pictures for forward reference pictures to be referred to when coding a P picture and a B picture, and coding modes for a B picture. That is, the moving picture coding apparatus 50 is provided with, instead of the control unit 110 and the mode selection unit 109 according to the first embodiment, a control unit 150 and a mode selection unit 159 which operate in different manners from those described for the first embodiment. To be specific, the control unit 150 according to the fifth embodiment controls a reference picture memory 117 in such a manner that, when coding a P picture, four pictures (I or P pictures) which are positioned forward the P picture are used as candidate pictures for forward reference, and when coding a B picture, four pictures (I or P pictures) which are positioned forward the B picture, a forward B picture that is closest to the B picture, and a backward I or P picture are used as candidate pictures. Further, when coding a block (target block) in a P picture, the mode selection unit 159 according to the fifth embodiment selects, as a coding mode for the target block, one from among the intra-picture coding, the inter-picture predictive coding using a motion vector, and the inter-picture predictive coding using no motion vector (a motion is treated as zero). When coding a block (target block) in a B picture, the mode selection unit 159 selects, as a coding mode for the target block, one from among the intra-picture coding, the inter-picture predictive coding using a forward motion vector, the inter-picture predictive coding using backward motion vector, and the inter- picture predictive coding using a forward motion vector and a backward motion vector. That is, the mode selection unit 159 of the moving picture coding apparatus 50 according to this fifth embodiment is different from the mode selection unit 109 of the moving picture coding apparatus 10 according to the first embodiment only in that it does not use the direct mode, and therefore, the moving picture coding apparatus 50 does not have the motion vector storage unit 116 of the moving picture coding apparatus 10. Further, the moving picture coding apparatus 50 according to the fifth embodiment is identical to the moving picture coding apparatus 10 according to the first embodiment except the coding control unit 150 and the mode selection unit 159. Next, the operation will be described. Input pictures are stored in the input picture memory 101, in units of pictures in order of display times. As shown in figure 29(a), input pictures P0, B1, B2, P3, B4, B5, P6, B7, B8, P9, B10, B11, P12, B13, B14, P15, B16, B17, and P18 are stored in the input picture memory 101 in order of display times. The respective pictures stored in the input picture memory 101 are rearranged in coding order as shown in figure 29(b). This rearrangement is carried out according to the relationships between target pictures and reference pictures during inter- picture predictive coding. That is, rearrangement of the input pictures is carried out such that a second picture to be used as a candidate for a reference picture when coding a first picture should be coded prior to the first picture. In this fifth embodiment, when coding a P picture (target picture), four pictures (I or P pictures) which are positioned timewise forward and close to the target picture are used as candidates for a reference picture. Further, when coding a B picture, four pictures (I or P pictures) which are positioned timewise forward and close to the target picture, a B picture which is positioned timewise forward and closest to the target picture, and an I or P picture which is positioned timewise backward and closest to the target picture, are used as candidates for a reference picture. The respective pictures rearranged in the input picture memory 101 are read out for each unit of motion compensation. In this fifth embodiment, the unit of motion compensation is a rectangle area (macroblock) in which pixels are arranged in matrix, having a size of 16 pixels in the horizontal direction X 16 pixels in the vertical direction. In the following description, a macroblock is simply referred to as a block. Hereinafter, coding processes for the pictures P15, B13, and B14 will be described in this order. (Coding Process for Picture P15) Since the picture P15 is a P picture, this picture is subjected to inter-picture predictive coding using forward reference. Further, in coding a P picture, no B picture is used as a reference picture. Figure 28 shows the manner of picture management in the reference picture memory 117. For example, at start of coding the picture P15, in the reference picture memory 117, the pictures P12, B11, P9, P6, and P3 are stored in memory areas to which logical memory numbers are assigned, in ascending order of the logical memory numbers. These pictures have already been coded, and the image data stored in the reference picture memory 117 are image data which have been decoded in the moving picture coding apparatus 50. Hereinafter, for simplification, a picture whose image data is stored in the memory is referred to as a picture stored in the memory. The reference candidate pictures stored in the reference picture memory 117 are assigned reference picture indices under control of the coding control unit 150. The assignment of reference picture indices is carried out not in order of picture coding but in order of display times. To be specific, a smaller reference picture index is assigned to a newer reference candidate picture, i.e., a reference candidate picture which is later in display order. However, in coding a P picture, no reference picture indices are assigned to B pictures. Further, in coding a B picture, a newest reference candidate picture is assigned a code [b] indicating that this picture should be treated as a backward reference picture. According to the above-mentioned reference picture index determining method, as shown in figure 28, reference picture indices [0], [1], [2], and [3] are assigned to the pictures P12, P9, P6, and P3, respectively, and no reference picture index is assigned to the picture B11. By the way, in coding a P picture, the coding control unit 150 controls the respective switches so that the switches 113, 114, and 115 are turned ON. A block in the picture P15 that is read from the input picture memory 101 is input to the motion vector detection unit 108, the mode selection unit 109, and the difference calculation unit 102. The motion vector detection unit 108 detects a motion vector of the block in the picture P15, using the pictures P12, P9, P6, and P3 to which the reference picture indices are assigned, among the pictures stored in the input picture memory 117. In this case, an optimum reference candidate picture is selected from among the pictures P12, P9, P6, and P3, and detection of the motion vector is carried out with reference to the selected reference picture. Then, the detected motion vector is output to the mode selection unit 159 and the bit stream generation unit 104. Further, information Rp indicating which one of the pictures P12, P9, P6, and P3 is referred to in detecting the motion vector, i.e., the reference picture index, is also output to the mode selection unit 159. The mode selection unit 159 determines a coding mode for the block in the picture P15, using the motion vector detected by the motion vector detection unit 108. The coding mode indicates a method for coding the block. For example, for a block in a P picture, a coding mode is selected from among the intra-picture coding, the inter-picture predictive coding using a motion vector, and the inter-picture predictive coding using no motion vector (i.e., motion is regarded as 0). Generally, selection of a coding mode is carried out so that coding error at a predetermined amount of bits is minimized. The coding mode Ms determined by the mode selection unit 159 is output to the bit stream generation unit 104. Further, when the determined coding mode is the coding mode which performs forward reference, the reference picture index is also output to the bit stream generation unit 104. Further, a prediction image Pd which is obtained on the basis of the coding mode determined by the mode selection unit 152 is output to the difference calculation unit 102 and the addition unit 106. However, when the intra-picture coding is selected, no prediction image Pd is outputted. Further, when the intra-picture coding is selected, the switch 111 is controlled so that the input terminal Ta is connected to the output terminal Tb2, and the switch 112 is controlled so that the output terminal Td is connected to the input terminal Tc2. Hereinafter, a description will be given of a case where the inter-picture predictive coding is selected in the mode selection unit 109. Since the operations of the difference calculation . unit 102, prediction error coding unit 103, bit stream generation unit 104, and prediction error decoding unit 105 are identical to those mentioned for the first embodiment, repeated description is not necessary. When coding of all blocks in the picture P15 is completed, the coding control unit 150 updates the logical memory numbers and the reference picture indices corresponding to the pictures stored in the reference picture memory 117. That is, since the coded picture P15 is later in the order of display times than any pictures stored in the reference picture memory 117, the picture P15 is stored in the memory area in which the logical memory number (0) is set. Then, the logical memory numbers of the memory areas where other reference pictures have already been stored are incremented by 1. Further, since the next target picture to be coded is the picture B13 that is a B picture, a reference picture index is also assigned to the picture B11. Thereby, the pictures P15, P12, B11, P9, P6, and P3 are stored in the memory areas in which the logical memory numbers (0)~(5) are set, respectively, and the reference picture indices [0], [1], [2], [3], and [4] are assigned to the pictures P12, B11, P9, P6, and P3, respectively. Since the next target picture is a B picture, the picture P15 stored in the memory area of the logical memory number 0 is assigned a code [b] indicating that this picture is treated as a backward reference picture, instead of the reference picture index. (Coding Process for Picture B13> Since the picture B13 is a B picture, this picture is subjected to inter-picture predictive coding using bidirectional reference. In this -case, four I or P pictures which are timewise close to the target picture and a B picture which is timewise closest to the target picture are used as candidate pictures for forward reference, and an I or P picture which is timewise closest to the target picture is used as a candidate picture for backward reference. Accordingly, the candidate pictures for forward reference for the picture B13 are the pictures P12, B11, P9, P6, and P3, and the candidate picture for backward reference for the picture B13 is the picture P15. These reference candidate pictures are stored in the reference picture memory 117. These reference candidate pictures are assigned logical memory numbers and reference picture indices as shown in figure 28. In coding a B picture, the coding control unit 150 controls the respective switches so that the switches 113, 114, and 115 are turned ON. Accordingly, a block in the picture B11 that is read from the input picture memory 101 is input to the motion vector detection unit 108, the mode selection unit 109, and the difference calculation unit 102. The motion vector detection unit 108 detects a forward motion vector and a backward motion vector of the block in the picture B13, using the pictures P12, B11, P9, P6, and P3 stored in the reference picture memory 117, as candidate pictures for forward reference, and the picture P15 as a candidate picture for backward reference. In this case, an optimum picture is selected from among the pictures P12, B11, P9, P6, and P3, and detection of the forward motion vector is carried out with reference to the selected picture. Then, the detected motion vector is output to the mode selection unit 159 and the bit stream generation unit 104. Further, information Rp indicating which one of the pictures P12, B11, P9, P6, and P3 is referred to in detecting the forward motion vector, i.e., the reference picture index, is also output to the mode selection unit 159. The operations of the mode selection unit 150, difference calculation unit 102, bit stream generation unit 104, and prediction error decoding unit 105 are identical to those for coding the picture P15. When coding of all blocks in the picture B13 is completed, the coding control unit 150 updates the logical memory numbers and the reference picture indices corresponding to the pictures stored in the reference picture memory 117. That is, since the picture B13 is positioned, in order of display times, before the picture P15 stored in the reference picture memory 117 and after the picture P12 stored in the reference picture memory 17, the picture B13 is stored in the memory area in which the logical memory number (1) is set. Further, since the picture B11 is not used as a reference picture in coding the subsequent pictures, the picture B11 is deleted. . At this time, information indicating that the picture B11 is deleted from the reference picture memory is output to the bit stream generation unit 104 as a control signal Cs1. The bit stream generation unit 104 describes this information as header information in the bit stream. Further, the logical memory number of the memory area corresponding to the picture P12 is incremented by 1. The next target picture to be coded is the picture B14 as a B picture, Accordingly, the picture stored in the memory area with the logical memory number (0) is used as a backward reference picture, and reference picture indices are assigned to the other pictures. Thereby, the pictures P15, B13, P12, P9, P6, and P3 are stored in the memory areas corresponding to the logical memory numbers (0)~(5), respectively, and the reference picture indices [0], [1], [2], [3], and [4] are assigned to the pictures B13, P12, P9, P6, and P3, respectively. (Coding Process for Picture B14) Since the picture B14 is a B picture, this picture is subjected to inter-picture predictive coding using bidirectional reference. As reference pictures for the picture B14, the pictures B13, P12, P9, P6, and P3 are used as forward reference pictures while the picture P15 is used as a backward reference picture. In processing a B picture, the coding control unit 150 controls the respective switches so that the switches 113, 114, and 115 are turned ON. Accordingly, a block in the picture B14 that is read from the input picture memory 101 is input to the motion vector detection unit 108, the mode selection unit 109, and the difference calculation unit 102. The motion vector detection unit 108 detects a forward motion vector and a backward motion vector of the block in the picture B14, using the pictures B13, P12, P9, P6, and P3 stored in the reference picture memory 117 as candidate pictures for forward reference as well as the picture P15 as a candidate picture for backward reference. In this case, an optimum picture is selected from among the pictures B13, P12, P9, P6, and P3, and detection of the forward motion vector is carried out with reference to the selected picture. Then, the detected motion vector is output to the mode selection unit 159 and the bit stream generation unit 104. Further, information Rp indicating which one of the pictures B13, P12, P9, P6, and P3 is referred to in detecting the forward motion vector, i.e., the reference picture index, is also output to the mode selection unit 159. The operations of the mode selection unit 150, difference calculation unit 102, bit stream generation unit 104, prediction error decoding unit 105, and addition unit 106 are similar to those for coding the picture P15. When coding of all blocks in the picture B14 is completed, the coding control unit 150 updates the logical memory numbers and the reference picture indices corresponding to the pictures stored in the reference picture memory 117. That is, since the picture B14 is positioned, in order of display times, before the picture P15 stored in the reference picture memory 117, and later than the picture B13 stored in the reference picture memory 117, the picture B14 is stored in the memory area in which the logical memory number (1) is set. Further, since the picture B13 is not used as a reference picture in coding the subsequent pictures, the picture B13 is deleted. At this time, information indicating that the picture B13 is deleted from the reference picture memory is output to the bit stream generation unit 104 as a control signal Cd1. The bit stream generation unit 104 describes this information as header information in the bit stream. The next target picture to be coded is the picture P18 that is a P picture. Accordingly, reference picture indices are assigned to the pictures other than B pictures. Thereby, the pictures P15, B14, P12, P9, and P6 are stored in the memory areas corresponding to the logical memory numbers (0)~(5), respectively, and the reference picture indices [0], [1], [2], and [3] are assigned to the pictures P15, P12, P9, and P6, respectively. As described above, according to the fifth embodiment, plural candidate pictures for forward reference for a target picture to be coded are assigned reference picture indices such that a smaller index is assigned to a candidate picture whose display time is later (i.e., information for identifying which one of the candidate pictures is used in detecting the forward motion vector of the target block). Therefore, a candidate picture which is most likely to be selected as a reference picture among the plural candidate pictures is assigned a smaller reference picture index. Accordingly, the amount of codes of the reference picture indices can be minimized, resulting in an increase in coding efficiency. Hereinafter, the effects of this fifth embodiment will be described taking a case where coding of a B picture is carried out using another B pictures as a reference candidate picture, together with the problems of the prior art. For example, it is assumed that pictures of a moving picture are arranged in display order as shown in figure 2 9(a), and four P pictures and one B picture are used as candidate pictures for forward reference in coding a target picture. Figure 30 shows an example of management of pictures stored in the reference picture memory. The candidate pictures are stored in coding order, in the memory. When coding the picture P15, in the reference picture memory, the pictures B11, P12, P9, P6, and P3 are stored in the memory areas, in ascending order of the logical memory numbers. Further these candidate pictures are assigned the reference picture indices [0], [1], [2], [3], and [4], respectively. Therefore, a reference picture index is assigned to a B picture (picture B11 in this case) which is not used as a reference picture in coding a P picture, and the reference picture index not to be used causes degradation in coding efficiency. Further, when coding the picture B13, in the reference picture memory, the pictures P15, B11, P12, P9, P6, and P3 are stored in the memory areasr in ascending order of the logical memory numbers. The picture P15 is assigned a code [b] indicating that this picture is used as a backward reference picture, and the remaining pictures are assigned the reference picture indices [0], [1], [2], [3], and [4], respectively. Therefore, the reference picture index assigned to the picture B11 that is timewise far from the picture B13 (target picture) is smaller than the reference picture index assigned to the picture P12 that is timewise close to the target picture B13. In performing motion detection, generally, a candidate picture that is timewise closer to a target picture is more likely to be used as a reference picture. Accordingly, when the reference picture index of the picture B11 that is far from the target picture is smaller than the reference picture index of the picture P12 that is close to the target picture, coding efficiency is degraded. Furthermore, when coding the picture B14, in the reference picture memory, the pictures B13, P15, B11, P12, P9, and P6 are stored in the memory areas, in ascending order of the logical memory numbers. The picture B13 is assigned a code [b] indicating that this picture is used as a backward reference picture, and the remaining pictures are assigned the reference picture indices [0], [1], [2], [3], and [4], respectively. Therefore, the picture P15 which should actually be used as a candidate picture for backward reference for the picture B14, is used as a candidate picture for forward reference. Moreover, the picture B13 which should actually be used as a candidate picture for forward reference for the picture B14, is used as a candidate picture for backward reference. As a result, it becomes difficult to perform correct coding. Further, in coding the picture B14, the picture B11 which is not used as a reference picture exists in the reference picture memory. On the other hand, according to the fifth embodiment of the invention, as shown in figure 28, the reference candidate pictures for the target picture are stored in display order in the reference picture memory, and the candidate pictures for forward reference are assigned the reference picture indices such that a candidate picture whose display time is later is assigned a smaller reference picture index, and therefore, a candidate picture which is more likely to be selected as a reference picture from among the candidate pictures is assigned a smaller reference picture index. Thereby, the amount of codes of the reference picture indices can be minimized, resulting in an increase in coding efficiency. Further, since, in coding a P picture, no reference picture indices are assigned to B pictures, occurrence of reference picture indices that will never be used is avoided, resulting in a further increase in coding efficiency. Moreover, when coding a B picture, no reference picture index is assigned to the picture that is stored in the memory area corresponding to the smallest logical memory number, and this picture is used as a backward reference picture. Therefore, in predictive coding of a B picture, a P picture to be used as a backward reference picture is prevented from being used as a forward reference picture. Further, when a picture that is not used as a reference picture is deleted from the reference picture memory, information indicating this deletion is described in the bit stream. Therefore, the decoding end can detect that the picture which is not to be used as a reference picture in decoding a target picture and the following pictures, is deleted from the reference picture memory. In this fifth embodiment, motion compensation is performed in units of image spaces (macroblocks) each comprising 16 pixels in the horizontal direction X 16 pixels in the vertical direction, and coding of a prediction error image is performed in units of image spaces (subblocks) each comprising 8 pixels in the horizontal direction X 8 pixel's in the vertical direction. However, the number of pixels in each macroblock (subblock) in motion compensation (coding of a prediction error image) may be different from that described for the fifth embodiment. Further, while in this fifth embodiment the number of continuous B pictures is two, the number of continuous B pictures may be three or more. Further, while in this fifth embodiment four pictures are used as candidate pictures for a forward reference picture in coding a P picture, the number of forward reference candidate pictures for a P picture may be other than four. Furthermore, while in this fifth embodiment four P pictures and one B picture are used as candidate pictures for a forward reference picture in coding a B picture, forward reference candidate pictures for a B picture are not restricted thereto. Furthermore, in this fifth embodiment, each of plural pictures constituting a moving picture, which is a target picture to be coded, is used as a reference picture when coding another picture that follows the target picture. However, the plural pictures constituting a moving picture may include pictures not to be used as reference pictures. In this case, the pictures not to be used as reference pictures are not stored in the reference picture memory, whereby the same effects as described for the fifth embodiment can be achieved. Furthermore, while in this fifth embodiment coding of a B picture is carried out using another B picture as a reference / candidate picture, coding of a B picture may be carried out without referring to another B picture. In this case, no B pictures are stored in the reference picture memory. Also in this case, the same effects as described for the fifth embodiment can be achieved by assigning reference picture indices according to the order of picture display times. Furthermore, while in this fifth embodiment a single system of reference picture indices are assigned, different systems of reference picture indices may be assigned in the forward direction and the backward direction, respectively. Moreover, while in this fifth embodiment a smaller reference picture index is assigned to a candidate picture for forward reference whose display time is later, the reference picture index assignment method is not restricted thereto so long as a smaller reference picture index is assigned to a candidate picture that is more likely to be selected as a reference picture. Figure 31 is a conceptual diagram illustrating the structure of a bit stream (format of a coded image signal) corresponding to pictures to which reference picture indices are assigned. A coded signal Pt equivalent to one picture includes header information Hp placed at the beginning of the picture, and a data section Dp that follows the header information Hp. The header information Hp includes a control signal (RPSL). The data section Dp includes coded data (bit stream) corresponding to each block. For example, a bit stream BLx is a bit stream of a block that is coded in intra-picture coding mode, and a bit stream BLy is a bit stream of a block that is coded in inter-picture predictive coding mode other than intra-picture coding mode. The block bit stream BLx includes header information Hbx, information Prx relating to a coding mode, and coded image information Dbx. The block bit stream BLy includes header information Hby, information Pry relating to a coding mode, first reference picture index R1d1, a second reference picture index Rld2, a first motion vector MV1, a second motion vector MV2, and coded image information Dby. Which of the first and second reference picture indices Rldl and Rld2 should be used is determined from the information Pry relating to the coding mode. A reference picture index Rldl is assigned to a forward reference candidate picture with priority over a backward reference candidate picture. A reference picture index Rld2 is assigned to a backward reference candidate picture with priority over a forward reference candidate picture. [Embodiment 6] Figure 32 is a block diagram for explaining a moving picture decoding apparatus 60 according to a sixth embodiment of the present invention. The moving picture decoding apparatus 60 according to the sixth embodiment decodes the bit stream Bs outputted from the moving picture coding apparatus 50 according to the fifth embodiment. The moving picture decoding apparatus 60 is different from the moving picture decoding apparatus 20 according to the second embodiment in candidate pictures for forward reference pictures to be referred to when coding a P picture and a B picture, and coding modes for a B picture. That is, the moving picture decoding apparatus 60 is provided, with, instead of the memory control unit 204 and the mode decoding unit 223 according to the second embodiment, a memory control unit 264 and a mode decoding unit 2 63 which operate in different manners from those described for the second embodiment. To be specific, the memory control unit 2 64 according to the sixth embodiment controls a reference picture memory 207 such that, when decoding a P picture, four pictures (I or P pictures) which are positioned forward the P picture are used as candidate pictures for forward reference, and when decoding a B picture, four pictures (I or P pictures) which are positioned forward the B picture, a forward B picture that is closest to the B picture, and a backward I or P picture are used as candidate pictures. Further, when decoding a block (target block) in a P picture, the mode decoding unit 263 according to the sixth embodiment selects, as a coding mode for the target block, one from among plural modes as follows: intra-picture decoding, inter-picture predictive decoding using a motion vector, and inter-picture predictive decoding using no motion vector (a motion is treated as zero). When decoding a block (target block) in a B picture, the mode decoding unit 2 63 selects, as a decoding mode for the target block, one from among plural modes as follows: intra- picture decoding, inter-picture predictive decoding using a forward motion vector, inter-picture predictive decoding using backward motion vector, and inter-picture predictive decoding using a forward motion vector and a backward motion vector. That is, the mode decoding unit 263 of the moving picture decoding apparatus 60 according to this sixth embodiment is different from the mode decoding unit 223 of the moving picture decoding apparatus 20 according to the second embodiment only in that it does not use a decoding process corresponding to the direct mode, and therefore, the moving picture decoding apparatus 60 does not have the motion vector storage unit 226 of the moving picture decoding apparatus 20. Further, the moving picture decoding apparatus 60 according to the sixth embodiment is identical to the moving picture decoding apparatus 20 according to the second embodiment except the memory control unit 264 and the mode decoding unit 263. Next, the operation of the moving picture decoding apparatus 60 will be described. The bit stream Bs outputted from the moving picture coding apparatus 50 according to the fifth embodiment is input to the moving picture decoding apparatus 60 shown in figure 32. In the bit stream Bs, each P picture has been subjected to inter-picture predictive coding, using four I or P pictures which are positioned timewise forward and close to the P picture, as reference candidate pictures. Further, each B picture has been coded using four P pictures which are positioned timewise forward and closest to the B picture, a B picture which is positioned timewise forward the B picture, and an I or P picture which is positioned timewise backward and closest to the B picture. In this case, the order of the pictures in the bit stream is as shown in figure 29(b). Hereinafter, decoding processes for the pictures P15, B13, and B14 will be described in this order. The bit stream of the picture P15 is input to the bit stream analysis unit 201. The bit stream analysis unit 201 extracts various kinds of data from the inputted bit stream Bs. The various kinds of data are information such as a coding mode, a motion vector, and the like. The extracted information for mode selection (coding mode) Ms is output to the mode decoding unit 263. Further, the extracted motion vector MV is output to the motion compensation decoding unit 205. Furthermore, the prediction error coded data Ed is output to the prediction error decoding unit 202. The mode decoding unit 263 controls the switches 209 and 210 with reference to the coding mode Ms extracted from the bit stream. When the coding mode is inter-picture coding, the switch 209 is controlled such that the input terminal Te is connected to the output terminal Tf1, and the switch 210 is controlled such that the output terminal Th is connected to the input terminal Tg1. When the coding mode is inter-picture predictive coding, the switch 209 is controlled such that the input terminal Te is connected to the output terminal Tf1, and the switch 210 is controlled such that the output terminal Th is connected to the input terminal Tg2. Further, the mode decoding unit 263 outputs the coding mode Ms also to the motion compensation decoding unit 205. Hereinafter, a description will be given of the case where the coding mode is inter-picture predictive coding. The prediction error decoding unit 202 decodes the inputted coded data Ed to generate prediction error data PDd. The generated prediction error data PDd is output to the switch 209. Since the input terminal Te of the switch 209 is connected to the output terminal Tf1, the prediction error data PDd is output to the addition unit 208. The motion compensation decoding unit 205 generates a motion compensation image from the inputted information such as the motion vector. The information inputted to the motion compensation decoding unit 205 is the motion vector MV and the reference picture index Rp. The motion compensation decoding unit 205 obtains a motion compensation image (prediction image) from the reference picture memory 207, on the basis of the inputted information. The picture P15 has been coded using the pictures P12, P9, P6, and P3 as candidates for a reference picture, and these candidate pictures have already been decoded and are. stored in the reference picture memory 207. Figure 28 shows the pictures stored in the reference picture memory 207. As shown in figure 28, when decoding the picture P15, the pictures P12,. B11, P9, P6, and P3 are stored in the reference picture memory 207. The memory control unit 264 assigns reference picture indices to the reference candidate pictures stored in the reference picture memory 117. This assignment of reference picture indices is carried according to the order of picture display times such that a smaller reference picture index is assigned to a newer reference candidate picture. In decoding a P picture, no reference picture indices are assigned to B pictures. Accordingly, reference picture indices [0], [1], [2], and [3] are assigned to the pictures P12, P9, P6, and P3, respectively, and no reference picture index is assigned to the picture B11. The motion compensation decoding unit 205 determines which one of the pictures P12, P9, P6, and P3 is used as a reference picture when coding the target block, from the reference picture indices. Then, the motion compensation decoding unit 205 obtains a prediction image (prediction data Pd) from the reference picture memory 207 on the basis of the determined reference picture and the motion vector to generate a motion compensation image. The motion compensation image so generated is input to the addition unit 208. The addition unit 208 adds the prediction error data PDd and the motion compensation image to generate a decoded image (data Ad) . The decoded image so generated is output through the switch 210 to the reference picture memory 207. When all of the macroblocks in the picture P15 have been decoded, the memory control unit 264 updates the logical memory numbers and the reference picture indices corresponding to the pictures stored in the reference picture memory 207. At this time, since, in order of time, the picture P15 is later than any pictures stored in the reference picture memory 117, the picture P15 is stored in the memory area in which the logical memory number (0) is set. Then, the logical memory numbers of the memory areas where other reference pictures have already been stored are incremented by 1. Further, since the next target picture to be decoded is the picture B13, a reference picture index is assigned to the picture B11. Thereby, the pictures P15, P12, B11, P9, P6, and P3 are stored in the memory areas in which the logical memory numbers (0)~(5) are set, respectively, and the reference picture indices [0], [1], [2], [3], and [4] are assigned to the pictures P12, B11, P9, P6, and P3, respectively. Since the operations of the bit stream analysis unit 201, the mode decoding unit 203, and the prediction error decoding unit 202 are identical to those described for decoding of the picture P15, repeated description is not necessary. The motion compensation decoding unit 205 generates a motion compensation image from the inputted information such as the motion vector. The information inputted to the motion compensation decoding unit 205 is the motion vector and the reference picture index. The picture B13 has been coded using the pictures P12, B11, P9, P6, and P3 as candidate pictures for forward reference, and the picture P15 as a candidate picture for backward reference. At decoding of the picture B13, these candidate pictures have already been decoded and are stored in the reference picture memory 207. When the coding mode is forward predictive coding or bidirectional predictive coding, the motion compensation decoding unit 205 determines which one of the candidate pictures P12, B11, P9, P6, and P3 is used as a forward reference picture when coding the picture B13, on the basis of the reference picture indices. Then, the motion compensation decoding unit 205 obtains a forward motion compensation image from the reference picture memory 207 on the basis of the determined reference picture and the motion vector. When the coding mode is bidirectional predictive coding or backward predictive coding, the motion compensation decoding unit 205 obtains a backward motion compensation image from the reference picture memory 207 on the basis of the determined reference picture and the backward motion vector. Then, the motion compensation decoding unit 2 05 generates a motion compensation image (prediction picture) using the forward motion compensation image and the backward motion compensation image. The motion compensation image so generated is output to the addition unit 208. The addition unit 208 adds the inputted prediction error image and motion compensation image to generate a decoded image. The decoded image so generated is output through the switch 210 to the reference picture memory 207. When all of the blocks in the picture B13 have been decoded, the memory control unit 264 updates the logical memory numbers and the reference picture indices corresponding to the pictures stored in the reference picture memory 207. Since the picture B13 is forward the picture P15 stored in the reference picture memory 207 in the order of display times and it is later than the picture P12 stored in the reference picture memory 207, the picture B13 is stored in the memory area in which the logical memory number (1) is set. Further, information indicating that the picture B11 is to be deleted from the reference picture memory is described in the bit stream, the memory control unit 264 controls the reference picture memory 207 so as to delete the picture B11 from the memory. Further, the logical memory number of the memory area where the other reference candidate picture P12 is stored is incremented by 1. Thereby, the pictures P15, B13, P12, P9, P6, and P3 are stored in the memory areas in which the logical memory numbers (0)~(5) are set, respectively, and the reference picture indices [0], [1], [2], [3], and [4] are assigned to the pictures B13, P12, P9, P6, and P3, respectively. Since the operations of the bit stream analysis unit 201, the mode decoding unit 203, and the prediction error decoding unit 202 are identical to those described for decoding of the picture P15, repeated description is not necessary. The motion compensation decoding unit 205 generates a motion compensation image from the inputted information such as the motion vector. The information inputted to the motion compensation decoding unit 205 is the motion vector and the reference picture index. The picture B14 has been coded using the pictures B13, P12, P9, P6, and P3 as candidate pictures for forward reference, and the picture P15 as a candidate picture for backward reference. At decoding of the picture B14, these candidate pictures have already been decoded and are stored in the reference picture memory 207. When the coding mode is forward predictive coding or bidirectional predictive coding, the motion compensation decoding unit 205 determines which one of the candidate pictures B13, P12, P9, P6, and P3 is used as a forward reference picture when coding the picture B14, on the basis of the reference picture indices. Then, the motion compensation decoding unit 205 obtains a forward motion compensation image from the reference picture memory 20 7 on the basis of the determined reference picture and the forward motion vector. When the coding mode is bidirectional predictive coding or backward predictive coding, the motion compensation decoding unit 205 obtains a backward motion compensation image from the reference picture memory 207 on the basis of the determined reference picture and the backward motion vector. Then, the motion compensation decoding unit 205 generates a motion compensation image, using the forward motion compensation image and the backward motion compensation image. The motion compensation image so generated is output to the addition unit 208. The addition unit 208 adds the inputted prediction error image and motion compensation image to generate a decoded image. The decoded image so generated is output through the switch 210 to the reference picture memory 207. When all of the blocks in the picture B14 have been decoded, the memory control unit 264 updates the logical memory numbers and the reference picture indices corresponding to the pictures stored in the reference picture memory 207. Since the picture B14 is forward the picture P15 stored in the reference picture memory 207 in the order of display times and it is later than the picture B13 stored in the input picture memory 207, the picture B14 is stored in the memory area in which the logical memory number (1) is set. Further, since information indicating that the picture B13 is to be deleted from the reference picture memory is described in the bit stream, the memory control unit 264 controls the reference picture memory 207 so as to delete the picture B13 from the memory. Since the next target picture to be decoded is the picture P18 that is a P picture, reference picture indices are assigned to pictures other than B pictures. Thereby, the pictures P15, B14, P12, P9, and P6 are stored in the memory areas in which the logical memory numbers (0)~(5) are set, respectively, and the reference picture indices [0], [1], [2], [3], and [4] are assigned to the pictures P15, P12, P9, and P6, respectively. Furthermore, the decoded pictures are outputted from the reference picture memory 207, as output images arranged in order of display times. Thereafter, the subsequent pictures are similarly decoded according to the picture type. As described above, according to the sixth embodiment, reference picture indices are assigned to plural candidate pictures for forward reference for a target picture to be decoded such that a smaller reference picture index is assigned to a candidate picture whose display time is later (i.e., information for identifying which candidate picture is referred to in detecting a forward motion vector of a target block), and a reference picture is determined from among the plural candidate pictures on the basis of the reference picture indices included in the bit stream of the target picture. Therefore, a smaller reference picture index is assigned to a candidate picture that is more likely to be used as a reference picture. Accordingly, it is possible to correctly decode a bit stream which is obtained by a highly-efficient coding method that can minimize the amount of codes corresponding to the reference picture indices. Further, since, in decoding a P picture, no reference picture indices are assigned to B pictures, it is possible to correctly decode a bit stream which is obtained by a highly- efficient coding method that can avoid occurrence of reference picture indices which will never be used. Furthermore, since, in decoding a B picture, a picture stored in a memory area on which a smallest logic memory number is set is used as a backward reference picture and no reference picture index is assigned to this picture, it is possible to correctly decode a bit stream which is obtained by a highly- efficient coding method that can prevent a P picture from being used as a forward reference picture in predictive coding of a B picture. Moreover, when information indicating that a picture which will never be used as a reference picture is deleted from the reference picture memory, is described in the bit stream, the reference picture is deleted from the reference picture memory according to the information, whereby the reference picture memory can be effectively used. Further, in this sixth embodiment, as an arrangement of plural pictures constituting a moving picture, an arrangement of pictures in which two B pictures are placed between adjacent P pictures. However, the number of B pictures pieced between adjacent P pictures may be other than two, for example, it may be three or four. Further, while in this sixth embodiment four pictures are used as candidate pictures for forward reference for a P picture, the number of forward reference candidate pictures for a P picture may be other than four. While in this sixth embodiment four P pictures and one B picture are used as candidate pictures for forward reference for a B picture, forward reference candidate pictures for a B picture are not restricted thereto. While in this sixth embodiment each of plural pictures constituting a moving picture is used as a reference picture when decoding another picture that follows this picture, plural pictures constituting a moving picture, which are to be decoded, may include pictures which will never be used as reference pictures. In this case, the pictures useless as reference pictures are not stored in the reference picture memory, whereby the same effects as described for the sixth embodiment can be achieved. While in this sixth embodiment decoding of a B picture is carried out using another B picture as a reference candidate picture, decoding of a B picture may be carried out without referring to another B picture. In this case, no B pictures are stored in the reference picture memory. Also in this case, the same effects as described for the sixth embodiment can be achieved by assigning reference picture indices according to the order of picture display times. While in this sixth embodiment, for simplification, a memory for managing reference candidate pictures, and a memory for rearranging decoded pictures in display order to output them are not separated but described as a single reference picture memory, the moving picture decoding apparatus 60 may be provided with a management memory for managing reference candidate pictures, and a rearrangement memory for rearranging decoded pictures in display order, respectively. In this case, the management memory is controlled by the memory controller 264, and outputs reference candidate pictures to the motion compensation decoding unit 205. Further, the rearrangement memory rearranges the decoded pictures arranged in decoding order, in display order, and sequentially outputs the pictures. Further, in this sixth embodiment, assignment of reference picture indices to candidate pictures is carried out according to a single rule, i.e., one system of reference picture indices are used. However, two systems of reference picture indices may be used, as described for the fifth embodiment. [Embodiment 7] Figure 33 is a block diagram for explaining a moving picture coding apparatus 70 according to a seventh embodiment of the present invention. This moving picture coding apparatus 70 is different from the moving picture coding apparatus iO according to the first embodiment in candidate pictures for forward reference pictures to be referred to when coding a P picture and a B picture, and coding modes for a B picture. That is, the moving picture coding apparatus 70 is provided with, instead of the control unit 110 and the mode selection unit 109 according to the first embodiment/ a coding control unit 170 and a mode selection unit 109 which operate in different manners from those described for the first embodiment. To be specific, the coding control unit 170 according to the seventh embodiment controls a reference picture memory 117 such that, when coding a P picture, three pictures (I or P pictures) which are positioned forward the P picture are used as candidate pictures for forward reference, and when coding a B picture, two pictures (I or P pictures) which are positioned forward the B picture, a forward B picture that is closest to the B picture, and a backward I or P picture are used as candidate pictures. However, a B picture, which is positioned forward an I or p picture that is positioned forward and closest to the target picture, is not referred to. The coding control unit 170 controls the bit stream generation unit 104 with a control signal Cd so that a flag indicating whether or not a target picture is to be referred to when coding subsequent pictures is inserted in the bit stream. To be specific, the code generation unit 104 is controlled with the control signal Cd so that information indicating that data of the target picture should be stored in the reference picture memory 117 at decoding as well as information indicating a period of time for the storage are added to the bit stream. Furthermore, when coding a block (target block) in a P picture, the mode selection unit 109 according to the seventh embodiment selects, as a coding mode for the target block, one from among plural modes as follows: intra-picture coding, inter- picture predictive coding using a motion vector, and inter- picture predictive coding using no motion vector (a motion is treated as zero). When coding a block (target block) in a B picture, the mode selection unit 179 selects, as a coding mode for the target block, one from among plural modes as follows: intra-picture coding, inter-picture predictive coding using a forward motion vector, inter-picture predictive coding using backward motion vector, and inter-picture predictive coding using a forward motion vector and a backward motion vector. That is, the mode selection unit 179 of the moving picture coding apparatus 70 according to this seventh embodiment is different from the mode selection unit 109 of the moving picture coding apparatus 10 according to the first embodiment only in that it does not use the direct mode, and therefore, the moving picture coding apparatus 10 does not have the motion vector storage unit 116 of the moving picture coding apparatus 10. Other constituents of the moving picture coding apparatus 70 according to the seventh embodiment are identical to those of the moving picture coding apparatus 10 according to the first embodiment. The moving picture coding apparatus 70 according to the seventh embodiment is different from the moving picture decoding apparatus 50 according to the fifth embodiment in that the coding control unit 170 controls the bit stream generation unit 104 so that a flag indicating whether or not a target picture is to be referred to when coding subsequent pictures is inserted in the bit stream. To be specific, the code generation unit 104 is controlled with the control signal Cd so that a flag indicating whether or not a target picture is to be referred to when coding subsequent pictures is inserted in the bit stream corresponding to the target picture. Further, the moving picture coding apparatus 70 is different from the moving picture coding apparatus 50 in candidate pictures to be referred to in coding a P picture and a B picture. The moving picture coding apparatus 70 is identical to the moving picture coding apparatus 50 in aspects other than those mentioned above. Next, the operation of the moving picture coding apparatus 70 will be described. Input image data Id are stored into the input picture memory 101, in units of pictures, in order of time. Figure 34(a) shows the order of pictures inputted to the input picture memory 101. As shown in figure 34(a), the respective pictures are successively inputted to the input picture memory 101, starting from a picture P1. In figure 34(a), pictures P1, P4, P7, P10, P13, P16, P19, and P22 are P pictures while pictures B2, B3, B5, B6, B8, B9, B11, B12, B14, P15, B17, P18, B20, and B21 are B pictures. When coding a P picture, three pictures (I or P pictures) which are timewise forward and close to the P picture are used as candidates for a reference picture. Further, when coding a B picture, two pictures (I or P pictures) which are timewise forward and close to the B picture, one B picture that is forward and closest to the B picture, and an I or P picture that is forward the B picture, are used as candidates for a reference picture. However, in coding a B picture, a B picture which is positioned forward an I or P picture that is timewise forward and closest to the B picture is not referred to. When coding an I picture, other pictures are not referred to. The data Id of the respective pictures inputted to the input picture memory 101 are rearranged in coding order. Thereinafter the data of each picture is referred to simply as a picture. That is, the process of changing the order of the pictures from input order to coding order is carried out on the basis of the relationships between target pictures and reference pictures in inter-picture predictive coding. In the rearrangement, the respective pictures are rearranged so that a second picture to be used as a candidate for a reference picture in coding a first picture is coded prior to the first picture. To be specific, the correspondences between the pictures P10 ~P13 and the reference candidate pictures are shown by arrows in figure 34(a). That is, when coding the P picture P10, the pictures P1, P4, and P7 are referred to, and when coding the P picture P13, the pictures P4, P7, and P10 are referred to. Further, when coding the B picture B11, the pictures P7, P10, and P13 are referred to, and when coding the B picture B12, the pictures P7, P10, B11, and P13 are referred to. Figure 34(b) shows the order of the pictures after rearranging the pictures B2 to P22 shown in figure 34(a). After the rearrangement, the respective pictures are arranged in order of P4, B2, B3, P7, B5, B6, P10, B8, B9, P13, B11, B12, P16, B14, B15, P19, B17, B18, and p22. The respective pictures rearranged in the reference picture memory 101 are successively read out, for each predetermined data processing unit, in order of coding times. In this seventh embodiment, the data processing unit is a unit of data on which motion compensation is carried out and, more specifically, it is a rectangle image space (macroblock) in which 16 pixels are arranged in both the horizontal direction and the vertical direction. In the following description, a macroblock is simply referred to as a block. Hereinafter, coding processes for the pictures P13, B11, and B12 will be described in this order. Since the picture P13 is a P picture, inter-picture predictive coding using forward reference is carried out as a coding process for the picture P13. In this case, three I or P pictures which are positioned forward the target picture (picture P13) are used as reference candidate pictures, and specifically, the pictures P4, P7, and P10 are used. These reference candidate pictures have already been coded, and the corresponding to decoded image data Dd are stored in the reference picture memory 117. In coding a P picture, the coding control unit 170 controls the respective switches so that the switches 113, 114, and 115 are turned ON. Data Md corresponding to a block in the picture P13, which is read from the input picture memory 101, is input to the motion vector detection unit 108, the mode selection unit 179, and the difference calculation unit 102. The motion vector detection unit 108 detects the motion vector MV of the block in the picture P13, using the decoded image data Rd of the pictures P4, P7, and P10 stored in the reference picture memory 117. In this case, an optimum picture is selected from among the pictures P4 P7, and P10, and detection of the motion vector is carried out with reference to the selected picture. Then, the detected motion vector MV is output to the mode selection unit 179 and the bit stream generation unit 104. Further, information indicating which one of the pictures P4, P7, and P10 is referred to in detecting the motion vector MV (reference picture information) is also output to the mode selection unit 179. The mode selection unit 179 determines a coding mode for the block in the picture P13, using the motion vector detected by the motion vector detection unit 108. To be specific, in the case of coding a P picture, a coding mode is selected from among the following coding modes: intra- picture coding, inter-picture predictive coding using a motion vector, and an inter-picture predictive coding using no motion vector (i.e., motion is regarded as 0). In determining a coding mode, generally, a coding mode which minimizes coding errors when a predetermined amount of bits is given to the block as an amount of codes, is selected. The coding mode Ms determined by the mode selection unit 179 is output to the bit stream generation unit 104. Further, when the determined coding mode Ms is the coding mode which performs forward reference, information indicating which one of the pictures P4, P7, and P10 is referred to in detecting the forward motion vector (forward motion vector) is also output to the bit stream generation unit 104. Then, prediction image data Pd, which is obtained from the reference picture according to the coding mode Ms that is determined by the mode selection unit 179, is output to the difference calculation unit 102 and the addition unit 106. However, when the intra-picture coding mode -is selected, no prediction image data Pd is outputted. Further, when the intra- picture coding is selected, the switches 111 and 112 are controlled in the same manner as described for the fifth embodiment. Hereinafter, a description will be given of a case where the inter-picture predictive coding mode is selected as the coding mode Ms. The difference calculation unit 102, the prediction error coding unit 103, the bit stream generation unit 104, the prediction error decoding unit 105, and the coding control unit 170 are identical to those described for the fifth embodiment. However, in this seventh embodiment, information indicating that the picture P13 is coded using forward three I or P pictures as reference candidate pictures, is added as header information of the picture P13. Further, since the picture P13 will be referred to when coding another picture, information (flag) indicating that decoded data Dd corresponding to the picture P13 should be stored in the reference picture memory 117 at decoding, is also added as header information of the picture P13. Further, information indicating that the picture P13 should be stored in the reference picture memory until decoding of the picture P22 is completed, is also added as header information of the picture P13. The storage period for the picture P13 may be indicated by time information of the picture P22 (e.g., time-basis positional information such as a picture number, decoding time information, or display time information), or period information from the picture P13 to the picture P22 (e.g., the number of pictures). The header information described above may be described as header information in picture units, i.e., as header information for every target picture to be coded. Alternatively, it may be described as header information of the entire sequence, or as header information in units of frames (e.g., in units of GOPs in MPEG) . When the coding mode for each block in the picture P13 is one performing forward reference, information indicating which one of the pictures P4, P7, and P10 is referred to in detecting the forward motion vector (reference picture information) is also added to the bit stream. For example, when the motion vector is obtained with reference to the picture P10, information indicating that the P picture just previous to the target picture is used as a reference picture (reference picture index) is added to the bit stream. When the motion vector is obtained with reference to the picture P7, information indicating that the P picture two-pictures previous to the target picture is used as a reference picture (reference picture index) is added to the bit stream. When the motion vector is obtained with reference to the picture P4, information indicating that the P picture three- pictures previous to the target picture is used as a reference picture (reference picture index) is added to the bit stream. For example, a reference picture index [0] may be used to indicate that the P picture just previous to the target picture is used as a reference picture, a reference picture index [1] may- be used to indicate that the P picture two-pictures previous to the target picture is used as a reference picture, and a reference picture index [2] may be used to indicate that the P picture three-pictures previous to the target picture is used as a reference picture. Further, information indicating that the P picture is subjected to inter-picture predictive coding using three reference candidate pictures is described as header information. The remaining macroblocks in the picture P13 are coded in like manner as described above. When all of the macroblocks in the picture P13 have been coded, coding of the picture B11 takes place. Since the picture B11 is a B picture, inter-picture predictive coding using bidirectional reference is carried out as a coding process for the picture B11. In this case, two pictures (I or P pictures which are timewise close to the target picture (picture B11) and a B picture which is timewise closest to the target picture are used as candidate pictures for forward reference, and an I or P picture which is timewise closest to the target picture is used as a candidate picture for backward reference. However, a B picture which is positioned beyond an I or P picture closest to the target picture is never be referred to. Accordingly, the pictures P7 and P10 are used as forward reference pictures for the picture B11, and the picture P13 is used as a backward reference picture for the picture B11. In processing a first B picture between continuous two B pictures, since this first B picture is used as a reference picture in coding the other B picture, the coding control unit 17 0 controls the respective switches so that the switches 113, 114, and 115 are turned ON. Accordingly, the image data Md corresponding to the block in the picture B11, which is read from the input picture memory 101, is input to the motion vector detection unit 108, the mode selection unit 179, and the difference calculation unit 102. The motion vector detection unit 108 detects a forward motion vector and a backward motion vector corresponding to the target block in the picture B11, with reference to the pictures P7 and P10 stored in the reference picture memory 117, as candidate pictures for forward reference, and the picture P13 stored in the reference picture memory 117, as a backward reference picture. In this case, either the picture P7 or the picture P10 is selected as a most suitable reference picture, and detection of a forward motion vector is carried out according to the selected picture. The detected motion vectors are output to the mode selection unit 179 and the bit stream generation unit 104. Further, information indicating which one of the pictures P7 and P10 is referred to in detecting the forward motion vector (reference picture information) is also output to the mode selection unit 179. The mode selection unit 179 determines a coding mode for the target block in the picture B11, using the motion vectors detected by the motion vector detection unit 108. As a coding mode for the B picture, one of the following coding modes is selected: intra-picture coding mode, inter-picture predictive coding mode using a forward motion vector, inter-picture predictive coding mode using a backward motion picture, and inter-picture predictive coding mode using bidirectional motion vectors. Also in this case, a general method (mode) which minimizes coding errors when a predetermined amount of bits are given as the amount of codes, should be selected. The coding mode determined by. the mode selection unit 179 is output to the bit stream generation unit 104. Further, prediction image data Pd, which is obtained from the reference picture according to the coding mode Ms that is determined by the mode selection unit 179, is output to the difference calculation unit 102 and the addition unit 106. However, when the intra- picture coding mode is selected by the mode selection unit 179, no prediction image data Pd is outputted. Further, when the intra-picture coding is selected, the switches 111 and 112 are controlled in the same manner as described for the coding process of the picture P13. Hereinafter, a description will be given of a case where the inter-picture predictive coding is selected by the mode selection unit 179. In this case, the operations of the difference calculation unit 102, the prediction error coding unit 103, the bit stream generation unit 104, the prediction error decoding unit 105, and the coding control unit 170 are identical to those described for the fifth embodiment. When the coding mode is one performing forward reference, information indicating which one of the pictures P7 and P10 is referred to in detecting the forward motion vector (reference picture information) is also added to the bit stream. For. example, when picture P10 is referred to, reference picture information indicating that a candidate picture just previous to the target picture is used as a reference picture is added to the bit stream. When the picture P7 is referred to, reference picture information indicating that a candidate picture two- pictures previous to the target picture is used as a reference picture is added to the bit stream. For example, a reference picture index [0] may be used to indicate that a candidate picture just previous to the target picture is used as a reference picture, and a reference picture index [1] may be used to indicate that a candidate picture two-pictures previous to the target picture is used as a reference picture. Further, in this case, information indicating that the target B picture is subjected to inter-picture predictive coding using a forward B picture as a reference picture is not added as header information. Furthermore, information indicating that the forward reference candidate pictures for the target B picture are two I or P pictures and one B picture is added as header information. Moreover, information indicating that a B picture, which is positioned forward an I or P picture that is positioned forward and closest to the target B picture, is not referred to is added as header information. Thereby, it is possible to know the capacity of a reference picture memory that is needed in decoding the bit stream Bs generated in the moving picture coding apparatus 70 according to the seventh embodiment. The header information described above may be described as header information in units of pictures, i.e., as header information for every target picture to be coded. Alternatively, it may be described as header information of the entire sequence, or as header information in units of several pictures (e.g., in units of GOPs in MPEG). Further, since the picture B11 is used as a reference picture when coding a picture positioned backward the picture B11, information indicating that decoded image data Dd corresponding to the picture B11 should be stored in the reference picture memory 117 at decoding, is also added as header information. Further, information indicating that the data Dd should be stored in the reference picture memory 117 until decoding of the picture B12 is completed, is also added as header information. When all of the remaining blocks in the picture B11 have been coded, coding of the picture B12 takes place. (Coding Process for Picture B12) Since the picture B12 is a B picture, inter-picture predictive coding using bidirectional reference is carried out as a coding process for the picture B12. In this case, two I or P pictures which are timewise close to the target picture B12, and a B picture which is timewise closest to the target picture B12 are used as candidate pictures for forward reference. Further, an I or P picture which is timewise closest to the target picture B12 is used as a candidate picture for backward reference. To be specific, the pictures P7, P10, and B11 are used as candidate pictures for forward reference for the picture B12, and the picture P13 is used as a backward reference picture for the picture B12. Since the picture B12 is not used as a reference picture when coding another picture, the coding control unit 170 controls the respective switches with the control signal Cs1 so that the switch 113 is turned ON and the switches 114 and 115 are turned OFF. Accordingly, the image data Md corresponding to the block in the picture B12, which is read from the input picture memory 101, is input to the motion vector detection unit 108, the mode selection unit 179, and the difference calculation unit 102. The motion vector detection unit 108 detects a forward motion vector and a backward motion vector corresponding to the macroblock in the picture B12, with reference to the pictures P7, P10, and B11 stored in the reference picture memory 117, as forward reference pictures, and the picture P13 stored in the reference picture memory 117, as a backward reference picture. In this case, a most suitable reference picture is selected from among the pictures P7, P10, and B11, and detection of a forward motion vector is carried out according to the selected picture. The detected motion vectors are output to the mode selection unit 179 and the bit stream generation unit 104. Further, information indicating which one of the pictures P7, P10, and B11 is referred to in detecting the forward motion vector (reference picture information) is also output to the mode selection unit 179. The mode selection unit 179 determines a coding mode for the block in the picture B12, using the motion vectors detected by the motion Vector detection unit 108. As a coding mode for the B picture, one of the following coding modes is selected: intra- picture coding mode, inter-picture predictive coding mode using a forward motion vector, inter-picture predictive coding mode using a backward motion picture, and inter-picture predictive coding mode using bidirectional motion vectors. The coding mode Ms determined by the mode selection unit 179 is output to the bit stream generation unit 104. Further, prediction image data Pd, which is obtained from the reference picture according to the coding mode that is determined by the A. mode selection unit 179, is output to the difference calculation unit 102 and the addition unit 106. However, when the intra- picture coding mode is selected, no prediction image data Pd is output. Further, when the intra-picture coding mode is selected by the mode selection unit 179, the switches 111 and 112 are controlled in the same manner as described for the coding process of the picture P13. Hereinafter, a description will be given of a case where the inter-picture predictive coding mode is selected by the mode selection unit 179. In this case, the operations of the difference calculation unit 102, the prediction error coding unit 103, the bit stream generation unit 104, the prediction error decoding unit 105, and the coding control unit 170 are identical to those described for the fifth embodiment. When the coding mode is one performing forward reference. information indicating which one of the pictures P7, P10, and B11 is referred to in detecting the forward motion vector (reference picture information) is also added to the bit stream. Further, information indicating that the target B picture B12 is subjected to inter-picture predictive coding using a forward B picture B11 as a candidate for a reference picture is described as header information. Furthermore, information indicating that the candidate pictures for forward reference are two I or P pictures and one B picture is described as header information. Moreover, information indicating that the picture B12 is not to be used as a reference picture when coding the following pictures is added as header information. Thereby, it is easily determined that there is no necessity to store the decoded image data Dd corresponding to the picture B12 in the reference picture memory at decoding, whereby management of the reference picture memory is facilitated. The above-mentioned header information may be described as header information in units of pictures, i.e., as header information for every target picture to be coded. Alternatively, it may be described as header information of the entire sequence, or as header information in units of several pictures (e.g., in units of GOPs in MPEG). The remaining blocks in the picture B12 are coded in the same manner as described above. Thereafter, the image data corresponding to the respective pictures following the picture B12 are coded in like manner as described above according to the picture type. For example, P pictures are processed like the picture P13, and the first B picture of the continuous B pictures (picture B14, B17, or the like) is processed like the picture P11. Further, the second B picture of the continuous B pictures (picture B15, B18, or the like) is processed like the picture P12. As described above, in the moving picture coding apparatus 70 according to the seventh embodiment, when coding a B picture as a target picture, since a B picture is also used as a candidate picture for forward reference as well as P pictures, a forward reference picture that is positioned closest to the target picture can be used as a forward reference picture. Thereby, prediction accuracy in motion compensation for a B picture can be increased, resulting in enhanced coding efficiency. Moreover, when coding a B picture as a target picture, information indicating whether or not the target picture is to be used as a reference picture when coding (decoding) another picture is added as header information. Further, when the target picture is used as a reference picture when coding (decoding) another picture, information indicating a period during which the target picture should be stored in the reference picture memory is added. Therefore, when decoding the bit stream Bs outputted from the moving picture coding apparatus 70, the decoding end can easily know which picture should be stored in the picture memory and how long the storage period is, whereby management of the reference picture memory at decoding is facilitated. In this seventh embodiment, when a target B picture is coded using another B picture as a reference picture, this is described as header information of the target B picture. However, the header information is not necessarily described in picture units. It may be described as header information of the entire sequence, or as header information in units of several pictures (e.g., GOP in MPEG) . In this seventh embodiment, motion compensation is performed in units of macroblocks each comprising 16 pixels (horizontal direction) X 16 pixels (vertical direction) , and coding of a prediction error image data is performed in units of blocks each comprising 4 pixels (horizontal direction) X 4 (vertical direction), or in units of blocks each comprising 8 pixels (horizontal direction) X 8 (vertical direction). However, motion compensation and coding of prediction error image data may be carried out in units of image spaces, each comprising different number of pixels from those mentioned above. Further, in this seventh embodiment, a coding mode for a P picture is selected from among intra-picture coding mode, inter- picture predictive coding mode using a motion vector, and inter- picture predictive coding mode using no motion vector, while a coding mode for a B picture is selected from among intra-picture coding mode, inter-picture predictive coding mode using a forward motion vector, inter-picture predictive coding mode using a backward motion vector, and inter-picture predictive coding mode using bidirectional motion vectors. However, selection of a coding mode for a P picture or a B picture is not restricted to that mentioned for the seventh embodiment. Further, while this seventh embodiment employs an image sequence in which two B pictures are inserted between an I picture and a P picture or between adjacent P pictures, the number of B pictures inserted between an I picture and a P picture or between adjacent P pictures in an image sequence may- be other than two, for example, it may be three or four. Furthermore, while in this seventh embodiment three pictures are used as candidate pictures for forward reference when coding a P picture, the number of forward reference candidate pictures for a P picture is not restricted thereto. Furthermore, while in this seventh embodiment two P pictures and one B picture are used as candidate pictures for forward reference when coding a B picture, forward reference candidate pictures to be used in coding a B picture are not restricted thereto. For example, forward reference candidate pictures for a B picture may be one P picture and two B pictures, or two P pictures and two B pictures, or three pictures which are timewise closest to the target picture regardless of the picture type. When, in coding a B picture, only one picture that is closest to the target B picture is used as a reference picture, it is not necessary to describe information indicating which picture is referred to in coding a target block in the B picture (reference picture information), in the bit stream. Further, in this seventh embodiment, when coding a B picture, a B picture which is positioned forward an I or P picture that is positioned forward and closest to the target B picture, is not referred to. However, when coding a B picture, a B picture which is positioned forward an I or P picture that is positioned forward and closest to the target B picture, may be used as a reference picture. [Embodiment 8] Figure 35 is a block diagram for explaining a moving picture decoding apparatus 80 according to an eighth embodiment of the present invention. The moving picture decoding apparatus 80 according to the eighth embodiment decodes the bit stream Bs outputted from the moving picture coding apparatus 70 according to the seventh embodiment. The moving picture decoding apparatus 80 is different from the moving picture decoding apparatus 20 according to the second embodiment in candidate pictures for forward reference pictures to be referred to when coding a P picture and a B picture, and decoding modes for a B picture. That is, the moving picture decoding apparatus 80 is provided with, instead of the memory control unit 204 and the mode decoding unit 223 according to the second embodiment, a memory control unit 284 and a mode decoding unit 283 which operate in different manners from those described for the second embodiment. To be specific, the memory control unit 284 according to the eighth embodiment controls a reference picture memory 287 such that, when decoding a P picture, three pictures (I or P pictures) which are positioned forward the P picture are used as candidate pictures for forward reference, and when decoding a B picture, two pictures (I or P pictures) which are positioned forward the B picture, a forward B picture that is closest to the B picture, and a backward I or P picture are used as candidate pictures. However, a B picture which is positioned forward an I or P picture that is positioned forward and closest to the target picture, is not referred to. The memory control unit 284 controls the reference picture memory 287, with a control signal Cm, on the basis of a flag indicating whether or not the target picture is to be referred to in coding a picture that follows the target picture, which flag is inserted in the code strong corresponding to the target picture. To be specific, information (flag) indicating that the data of the target picture should be stored in the reference picture memory 287 at decoding, and information indicating a period during which the data of the target picture should be stored, are included in the bit stream corresponding to the target picture. Further, when decoding a block (target block) in a P picture, the mode decoding unit 283 according to the eighth embodiment selects, as a coding mode for the target block, one from among the following modes: intra-picture decoding, inter-picture predictive decoding using a motion vector, and inter-picture predictive decoding using no motion vector (a motion is treated as zero). When decoding a block (target block) in a B picture, the mode decoding unit 283 selects, as a decoding mode for the target block, one from among the following modes: intra-picture decoding, inter-picture predictive decoding using a forward motion vector, inter-picture predictive decoding using backward motion vector, and inter-picture predictive decoding using a forward motion vector and a backward motion vector. That is, the mode decoding unit 283 of the moving picture decoding apparatus 80 according to this eighth embodiment is different from the mode decoding unit 223 of the moving picture decoding apparatus 20 according to the second embodiment only in that it does not use the direct mode, and therefore, the moving picture decoding apparatus 80 does not have the motion vector storage unit 226 of the moving picture decoding apparatus 20. Other constituents of the moving picture decoding apparatus 80 according to the seventh embodiment are identical to those of the moving picture decoding apparatus 20 according to the second embodiment. Further, the moving picture decoding apparatus 80 according to the eighth embodiment different from the moving picture decoding apparatus 60 according to the sixth embodiment in that the memory control unit 284 controls the bit stream generation unit 104 so that a flag indicating whether or not the target picture is to be referred to in coding a picture after the target block is inserted in the bit stream corresponding to the target picture. Further, in the moving picture decoding apparatus 80, candidate pictures to be referred to in decoding a P picture and a B picture are also different from those employed in the moving picture decoding apparatus according to the sixth embodiment. Other constituents of moving picture decoding apparatus 80 according to the seventh embodiment are identical to those of the moving picture decoding apparatus 60 according to the sixth embodiment. Next, the operation of the moving picture decoding apparatus 80 will be described. The bit stream Bs outputted from the moving picture coding apparatus 70 according to the seventh embodiment is input to the moving picture decoding apparatus 80. In this eighth embodiment, when decoding a P picture, three pictures (I or P pictures) which are timewise forward and close to the P picture are used as candidates for a reference picture. On the other hand, when decoding a B picture, two pictures (I or P pictures) which are positioned timewise forward and close to the B picture, a B picture which is positioned forward and closest to the B picture, and an I or P picture which is positioned backward the target picture, are used as candidate pictures for a reference picture. However, in decoding a B picture, a B picture which is positioned forward an I or P picture that is positioned forward and closest to the target picture, is not referred to. Further, in decoding an I picture, other pictures are not referred to. Further, information indicating which of the candidate pictures is used as a reference picture in decoding a P picture or a B picture is described as header information Ih of the bit stream Bs, and the header information Ih is extracted by the bit stream analysis unit 201. The header information Ih is output to the memory control unit 284. The header information may be described as header information of the entire sequence, header information in units of several pictures (e.g., GOP in MPEG), or header information in picture units. The pictures in the bit stream Bs inputted to the moving picture decoding apparatus 80 are arranged in order or picture decoding as shown in figure 36(a). Hereinafter, decoding processes for the pictures P13, B11, and B12 will be specifically described in this order. When the bit stream corresponding to the picture P13 is input to the bit stream analysis unit 201, the bit stream analysis unit 201 extracts various kinds of data from the inputted bit stream. The various kinds of data are information (coding mode) Ms relating to mode selection, information of the motion vector MV, the above-described header information, and the like. The extracted coding mode Ms is output to the mode decoding unit 283. Further, the extracted motion vector MV is output to the motion compensation decoding unit 205. Furthermore, the coded data Ed extracted by the bit stream analysis unit 201 is output to the prediction error decoding unit 202. The mode decoding unit 283 controls the switches 209 and 210 with reference to the mode selection information (coding mode) Ms extracted from the bit stream. When the coding mode Ms is intra- picture coding mode and when the coding mode Ms is inter-picture predictive coding mode, the switches 209 and 210 are controlled in like manner as described for the sixth embodiment. Further, the mode decoding unit 283 outputs the coding mode Ms to the motion compensation decoding unit 205. Hereinafter, a description will be given of the case where the coding mode-is inter-picture predictive coding mode. Since the operations of the prediction error decoding unit 202, the motion compensation decoding unit 205, and the addition unit 208 are identical to those described for the sixth embodiment, repeated description is not necessary. Figure 37 shows how the pictures, whose data are stored in the reference picture memory 2 07, change with time. When decoding of the picture P13 is started, the pictures B8, P7, and P10 are stored in areas R1, R2, and R3 of the reference picture memory 207. The picture P13 is decoded using the pictures P7 and P10 as candidates for a reference picture, and the picture P13 is stored in the memory area Rl where the picture B8 had been stored. Such rewriting of image data of each picture in the reference picture memory is carried out based on the header information of each picture which is added to the bit stream. This header information indicates that the picture P7 should be stored in the reference picture memory 207 until. decoding of the picture P13 is completed, the picture P10 should be stored in the memory until decoding of the picture P16 is completed, and the picture B8 should be stored in the memory until decoding of the picture B9 is completed. In other words, since it can be decided that the picture B8 is not necessary for decoding of the picture P13 and the following pictures, the picture P13 is written over the reference picture memory area Rl where the picture B8 is stored. Further, since information indicating that the picture P13 should be stored in the reference picture memory until decoding of the picture P19 is completed is described as header information of the picture P13, the picture P13 is stored in the reference picture memory at least until that time. As described above, the blocks in the picture P13 are successively decoded. When all of the coded data corresponding to the blocks in the picture P13 have been decoded, decoding of the picture B11 takes place. Since the operations of the bit stream analysis unit 201, the mode decoding unit 203, and the prediction error decoding unit 202 are identical to those described for decoding of the picture P13, repeated description is not necessary. The motion compensation decoding unit 205 generates motion compensation image data Pd from the inputted information such as the motion vector. That is, the information inputted to the motion compensation decoding unit 205 is the motion vector MV and reference picture index corresponding to the picture B11. The picture B11 has been coded using the picture P10 as a forward reference picture, and the picture P13 as a backward reference picture. Accordingly, in decoding of the picture B11, these candidate pictures P10 and P13 have already been decoded, and the corresponding decoded image data DId are stored in the reference picture memory 207. When the coding mode is bidirectional predictive coding mode, the motion compensation decoding unit 205 obtains a forward reference image from the reference picture memory 207 on the basis of the information indicating the forward motion vector, and obtains a backward reference image from the memory 207 on the basis of the information indicating the backward motion vector. Then, the motion compensation decoding unit 205 performs addition and averaging of the forward reference image and the backward reference image to generated a motion compensation image. Data Pd of the motion compensation image so generated is output to the addition unit 208. The addition unit 208 adds the inputted prediction error image data PDd and motion compensation image data Pd to output addition image data Ad. The addition image data Ad so generated is outputted as decoded image data DId, through the switch 210 to the reference picture memory 207. The memory control unit 284 controls the reference picture memory 207 on the basis of information indicating which candidate pictures are referred to in coding the P picture and the B picture, which information is header information of the bit stream. Figure 37 shows how the pictures stored in the reference picture memory 207 change with time. When decoding of the picture P11 is started, the pictures P13, P7, and P10 are stored in the reference picture memory 207. The picture Pll is decoded using the pictures P10 and P13 as reference pictures, and the picture P11 is stored in the memory area R2 where the picture P7 had been stored. Such rewriting of each picture in the reference picture memory 207 is carried out based on the header information of each picture which is added to the bit stream. This header information indicates that the picture P7 should be stored in the reference picture memory 20 7 until decoding of the picture P13 is completed, the picture P10 should be stored in the memory until decoding of the picture PI6 is completed, and the picture P13 should be stored in the memory until decoding of the picture P19 is completed. In other words, since it is decided that the picture P7 is not necessary for decoding of the picture P13 and the following pictures, the picture P11 is stored in the reference picture memory area R2 where the picture P7 is stored. Further, since information indicating that the picture B11 should be stored in the reference picture memory 207 until decoding of the picture B12 is completed is described as header information of the picture B11, the picture B11 is stored in the reference picture memory 207 at least until that time. As described above, the coded data corresponding to the blocks in the picture B11 are successively decoded. When all of the coded data corresponding to the blocks in the picture B11 have been decoded, decoding of the picture B12 takes place. (Decoding Process for Picture B12) Since the operations of the bit stream analysis unit 201, the mode decoding unit 203, and the prediction error decoding unit 202 are identical to those described for decoding of the picture P13, repeated description is not necessary. The motion compensation decoding unit 205 generates motion compensation image data Pd from the inputted information such as the motion vector. That is, the information inputted to the motion compensation decoding unit 205 is the motion vector MV and reference picture index corresponding to the picture B12. The picture B12 has been coded using the pictures P10 and B11 as candidates for a forward reference picture, and the picture P13 as a backward reference picture. These reference candidate pictures P10, B11, and P13 have already been decoded, and the corresponding decoded image data are stored in the reference picture memory 207. When the coding mode is bidirectional predictive coding mode, the motion compensation decoding unit 205 determined which one of the pictures P10 and B11 is used as a forward reference picture in coding the picture B12, according to the reference picture indices, and obtains a forward reference image from the reference picture memory 207 according to the information indicating the forward motion vector. Further, the motion compensation decoding unit 205 obtains a backward reference image from the memory 207 according to the information indicating the backward motion vector. Then, the motion compensation decoding unit 205 performs addition and averaging of the forward reference image and the backward reference image to generated a motion compensation image. Data Pd of the motion compensation image so generated is output to the addition unit 208. The addition unit 208 adds the inputted prediction error image data PDd and motion compensation image data Pd to output addition image data Ad. The addition image data Ad so generated is outputted as decoded image data DId, through the switch 210 to the reference picture memory 207. The memory control unit 284 controls the reference picture memory 207 on the basis of information indicating which reference pictures are used in coding the P picture and the B picture, v.hich information is extracted from the header information of the bit stream. Figure 37 shows how the pictures stored in the reference picture memory 207 change with time. When decoding of the picture B12 is started, the pictures P13, B11, and P10 are stored in the reference picture memory 207. The picture B12 is decoded using the pictures P13, B11, and P10 as reference candidate pictures. Since information indicating that the picture B12 is not to be used as a reference picture when decoding another picture is described as header information, the decoded data of the picture B12 is not stored in the reference picture memory 207 but outputted as output image data Od. As described above, the coded data corresponding to the blocks in the picture B12 are successively decoded. The decoded image data of the respective pictures which are stored in the reference picture memory 207, and the decoded image data which are not stored in the reference picture memory 207 are rearranged in order of their display times as shown in figure 36(b), and outputted as output image data Od. Thereafter, the coded data corresponding to the respective pictures are decoded in like manner as described above according to the picture type. To be specific, the coded data of the P pictures are decoded like the picture P13, and the first B picture (picture B14, B17, or the like) of the continuous B pictures is decoded like the picture P11. Further, the second B picture (picture B15, B18, or the like) of the continuous B pictures is decoded like the picture P12. As described above, in the moving picture decoding apparatus 80 according to the eighth embodiment, since a B picture is used as a reference candidate picture when decoding a B picture, a bit stream, which is obtained in a coding process that uses a B picture as well as P pictures as forward reference candidate pictures when coding a B picture, can be accurately decoded. Further, since the reference picture memory is controlled using information obtained from the bit stream, indicating which reference pictures are used in coding a P picture and a B picture, the reference picture memory can be effectively utilized. That is, image data of pictures to be used as reference pictures in the following decoding process are maintained in the reference picture memory, while image data of pictures not to be used as reference pictures in the following decoding process are successively erased from the memory, whereby the reference picture memory can be effectively utilized. While this eighth embodiment employs a bit stream corresponding to an image sequence in which two B pictures are inserted between adjacent P pictures, the number of B pictures positioned between adjacent P pictures may be other than two, for example, it may be three or four. Furthermore, while in this eighth embodiment two pictures are used as candidate pictures for forward reference when decoding a P picture, the number of forward reference candidate pictures to be referred to in decoding a P picture is not restricted thereto. Furthermore, in this eighth embodiment, when decoding a B picture, one P picture and one B picture are used as candidate pictures for forward reference, and a B picture which is positioned forward an I or P picture that is timewise closest to the target B picture, is not used as a reference picture. However, pictures to be used as reference candidate pictures in decoding a B picture may be other than those described for the eighth embodiment. Further, when decoding a B picture, a B picture which is positioned forward an I or P picture that is timewise closest to the target B picture, may be used as a reference picture. Furthermore, while in the eighth embodiment decoded image data of pictures which are not to be used as reference pictures when decoding other pictures are not stored in the reference picture memory, the decoded image data of these pictures may be stored in the memory. For example, when output of decoded image data of each picture is carried out with a little delay from decoding of each picture, the decoded image data of each picture must be stored in the reference picture memory. In this case, a memory area, other than the memory area where the decoded image data of the reference candidate pictures are stored, is provided in the reference picture memory, and the decoded image data of the pictures not to be used as reference pictures are stored in this memory area. Although, in this case, the storage capacity of the reference picture memory is increased, the method for managing the reference picture memory is identical to that described for the eighth embodiment and, therefore, the reference picture memory can be easily managed. While all pictures are used as reference candidate pictures in the second, fourth, sixth, and eighth embodiments, all pictures are not necessarily used as reference candidate pictures. To be brief, in a moving picture decoding apparatus, usually, already-decoded pictures are once stored in a decoding buffer (decoded frame memory) regardless of whether they will be used as reference candidate pictures or not, and thereafter, the already- decoded pictures are successively read from the decoding buffer to be displayed. In the second, fourth, sixth, and eighth embodiments of the present invention, all pictures are used as reference candidate pictures and, therefore, all of already-decoded pictures are stored in a reference picture memory for holding pictures to be used as reference candidate pictures, and thereafter, the already-decoded pictures are successively read from the reference picture memory to be displayed. However, as described above, all of the already-decoded pictures are not necessarily used as reference candidate pictures. Accordingly, the already-decoded pictures may be once stored in a decoding buffer (decoded frame memory) for holding not only pictures not to be used as reference candidate pictures but also pictures to be used as reference candidate pictures, and thereafter, the already-decoded pictures are successively read from the decoding buffer to be displayed. The moving picture coding apparatus or the moving picture decoding apparatus according to any of the aforementioned embodiments is implemented by hardware, while these apparatuses may be implemented by software. In this case, when a program for executing the coding or decoding process according to any of the aforementioned embodiments is recorded in a data storage medium such as a flexible disk, the moving picture coding apparatus or the moving picture decoding apparatus according to any of the aforementioned embodiments can be easily implemented in an independent computer system. Figures 38(a)-38(c) are diagrams for explaining a computer system for executing the moving picture coding process according to any of the first, third, fifth, and seventh embodiments and the moving picture decoding process according to any of the second, fourth, sixth, and eighth embodiments. Figure 38(a) shows a front view of a flexible disk FD which is a medium that contains a program employed in the computer system, a cross-sectional view thereof, and a flexible disk body D. Figure 38(b) shows an example of a physical format of the flexible disk body D. The flexible disk FD is composed of the flexible disk body D and a case FC that contains the flexible disk body D. On the surface of the disk body D, a plurality of tracks Tr are formed concentrically from the outer circumference of the disk toward the inner circumference. Each track is divided into 16 sectors Se in the angular direction. Therefore, in the flexible disk FD containing the above-mentioned program, data of the program for executing the moving picture coding process or the moving picture decoding process are recorded in the assigned storage areas (sectors) on the flexible disk body D. Figure 38(c) shows the structure for recording or reproducing the program in/from the flexible disk FD. When the program is recorded in the flexible disk FD, data of the program are written in the flexible disk FD from the computer system Csys through the flexible disk drive FDD. When the above-mentioned moving picture coding or decoding apparatus is constructed in the computer system Csys by the program recorded in the flexible disk FD, the program is read from the flexible disk FD by the flexible disk drive FDD and then loaded to the computer system Csys. Although in the above description a flexible disk is employed as a storage medium, an optical disk may be employed. Also in this case, the moving picture coding or decoding process can be performed by software in like manner as the case of using the flexible disk. The storage medium is not restricted to these disks, and any medium may be employed as long as it can contain the program, for example, a CD-ROM, a memory card, or a ROM cassette. Also when such data storage medium is employed, the moving picture coding or decoding process can be performed by the computer system in the same manner as the case of using the flexible disk. Applications of the moving picture coding method and the moving picture decoding method according to any of the aforementioned embodiments and systems using the same will be described hereinafter. Figure 39 is a block diagram illustrating an entire construction of a contents provision system 1100 that performs contents distribution services. A communication service provision area is divided into regions (cells) of desired size, and base stations. 1107 to 1110 which are each fixed radio stations are established in the respective cells. In this contents provision system 1100, various devices such as a computer 1111, a PDA (personal digital assistant) 1112, a camera 1113, a portable telephone 1114, and a portable telephone with a camera 1200 are connected to the Internet 1101 through an Internet service provider 1102, a telephone network 1104, and the base stations 1107 to 1110. However, the contents provision system 1100 is not restricted to a system including all of the plural devices shown in figure 39, but may be one including some of the plural devices shown in figure 39. Further, the respective devices may be connected directly to the telephone network 1104, not through the base stations 1107 to 1110 as the fixed radio stations. The camera 1113 is a device that can take moving pictures of an object, like a digital video camera. The portable telephone may be a portable telephone set according to any of PDC (Personal Digital Communications) system, CDMA (Code Division Multiple Access) system, W-CDMA (Wideband-Code Division Multiple Access) system, and GSM (Global System for Mobile Communications) system, or PHS (Personal Handyphone System). A streaming server 1103 is connected to the camera 1113 through the base station 1109 and the telephone network 1104. In this system, live distribution based on coded data which are transmitted by a user using the camera 1113 can be performed. The coding process for the data of taken images may be carried out by either the camera 1113 or the server that transmits the data. Moving picture data which are obtained by taking moving pictures of an object by means of the camera 1116 may be transmitted to the streaming server 1103 through the computer 1111. The camera 1116 is a device that can take still images or moving pictures of an object, such as a digital camera. In this case, coding of the moving picture data can be performed by either the camera 1116 or the computer 1111. Further, the coding process is carried out by an LSI 1117 included in the computer 1111 or the camera 1116. Image coding or decoding software may be stored in a storage medium (a CD-ROM, a flexible disk, a hard disk, or the like) which is a recording medium that contains data readable by the computer 1111 or the like. The moving picture data may be transmitted through the portable telephone with a camera 1200. The moving picture data are data which have been coded by an LSI included in the portable telephone 1200. In this contents provision system 1100, contents corresponding to images taken by the user by means of the camera 1113 or the camera 1116 (for example, live video of a music concert) are coded in the camera in the same manner as any of the aforementioned embodiments, and transmitted from the camera to the streaming server 1103. The contents data are subjected to streaming distribution from the streaming server 1103 to a requesting client. The client may be any of the computer 1111, the PDA 1112, the camera 1113, the portable telephone 1114 and the like, which can decode the coded data. In this contents provision system 1100, the coded data can be received and reproduced on the client side. When the data are received, decoded, and reproduced in real time on the client side, private broadcasting can be realized. The coding or decoding in the respective devices that constitute this system can be performed using the moving picture coding apparatus or the moving picture decoding apparatus according to any of the aforementioned embodiments. A portable telephone will be now described as an example of the moving picture coding or decoding apparatus. Figure 40 is a diagram illustrating a portable telephone 1200 that employs the moving picture coding method and the moving picture decoding method according to any of the aforementioned embodiments. This portable telephone 1200 includes an antenna 1201 for transmitting/receiving radio waves to/from the base station 1110, a camera unit 1203 that can take video or still images of an object, such as a CCD camera, and a display unit 1202 such as a liquid crystal display for displaying data of the video taken by the camera unit 1203 or video received through the antenna 1201. The portable telephone 1200 further includes a main body 1204 including plural control keys, a voice output unit 1208 for outputting voices such as a speaker, a voice input unit 1205 for inputting voices such as a microphone, a recording medium 1207 for retaining coded data or decoded data such as data of taken moving pictures or still images, or data, moving picture data or still image data of received e-mails, and a slot unit 1206 which enables the recording medium 1207 to be attached to the portable telephone 1200. The recording medium 1207 has a flash memory element as a type of EEPROM (Electrically Erasable and Programmable Read Only Memory) that is an electrically programmable and erasable non- volatile memory contained in a plastic case, like a SD card. The portable telephone 1200 will be described more specifically with reference to Figure 41. The portable telephone 1200 has a main control unit 1241 that performs general control for the respective units of the main body including the display unit 1202 and the control key 1204. The portable telephone 1200 further includes a power supply circuit 1240, an operation input control unit 1234, an image coding unit 1242, a camera interface unit 1233, a LCD (Liquid Crystal Display) control unit 1232, an image decoding unit 1239, a multiplexing/demultiplexing unit 1238, a recording/reproduction unit 1237, a modulation/demodulation unit 1236, and an audio processing unit 1235. The respective units of the portable telephone 1200 are connected to each other via a synchronization bus 1250. The power supply circuit 1240 supplies power from a battery pack to the respective units when a call end/power supply key is turned ON under the control of a user, thereby activating the digital portable telephone with a camera 1200 to be turned into an operable state. In the portable telephone 1200, the respective units operate under control of the main control unit 1241 that is constituted by a CPU, a ROM, a RAM and the like. To be more specific, in the portable telephone 1200, an audio signal that is obtained by voice inputting into the voice input unit 1205 in a voice communication mode is converted into digital audio data by the audio processing unit 1235. The digital audio data is subjected to a spectrum spread process by the modulation/demodulation circuit 1236, further subjected to a DA conversion process and a frequency transformation process by the transmission/receiving circuit 1231, and transmitted through the antenna 1201. In this portable telephone set 1200, a signal received through the antenna 1201 in the voice communication mode is amplified, and then subjected to a frequency transformation process and an AD conversion process. The received signal is further subjected to a spectrum inverse spread process in the modulation/demodulation circuit 1236, converted into an analog audio signal by the audio processing unit 1235, and this analog audio signal is outputted through the voice output unit 1208. When the portable telephone 1200 transmits an electronic mail in a data communication mode, text data of the e-mail that is inputted by manipulation of the control key 1204 on the main body is transmitted to the main control unit 1241 via the operation input control unit 1234. The main control unit 1241 controls the respective units so that the text data is subjected to the spectrum spread process in the modulation/demodulation circuit 1236, then subjected to the DA conversion process and the frequency transformation process in the transmission/receiving circuit 1231, and then transmitted to the base station 1110 through the antenna 1201. When this portable telephone 1200 transmits image data in the data communication mode, data of an image taken by the camera unit 1203 is supplied to the image coding unit 1242 via the camera interface unit 1233. When the portable telephone 1200 does not transmit the image data, the data of the image taken by the camera unit 1203 can be displayed directly on the display unit 1202 via the camera interface unit 1233 and the LCD control unit 1232. The image coding unit 1242 includes the moving picture coding apparatus according to any of the aforementioned embodiments. This image coding unit 1242 compressively encodes the image data supplied from the camera unit 1203 by the moving picture coding method according to any of the above embodiments to convert the same into coded image data, and outputs the obtained coded image data to the multiplexing/demultiplexing unit 1238. At the same time, the portable telephone 1200 transmits voices which are inputted to the voice input unit 1205 while the image is being taken by the camera unit 1203, as digital audio data, to the multiplexing/demultiplexing unit 1238 through the audio processing unit 1235. The multiplexing/demultiplexing unit 1238 multiplexes the coded image data supplied from the image coding unit 1242 and the audio data supplied from the audio processing unit 1235 by a predetermined method. Resultant multiplexed data is subjected to a spectrum spread process in the modulation/demodulation circuit 1236, then further subjected to the DA conversion process and the frequency transformation process in the transmission/receiving circuit 1231, and obtained data is transmitted through the antenna 1201. When the portable telephone 1200 receives data of a moving picture file that is linked to a home page or the like in the data communication mode, a signal received from the base station 1110 through the antenna 1201 is subjected to a spectrum inverse spread process by the modulation/demodulation circuit 1236, and resultant multiplexed data is transmitted to the multiplexing/demultiplexing unit 1238. When the multiplexed data that is received via the antenna 1201 is decoded, the multiplexing/demultiplexing unit 1238 demultiplexes the multiplexed data to divide the data into a coded bit stream corresponding to the image data and a coded bit stream corresponding to the audio data, and the coded image data is supplied to the image decoding unit 1239 and the audio data is supplied to the audio processing unit 1235, via the synchronization bus 1250. The image decoding unit 1239 includes the moving picture decoding apparatus according to any of the aforementioned embodiments. The image decoding unit 1239 decodes the coded bit stream of the image data by the decoding method corresponding to the coding method according to any of the above-mentioned embodiments, to reproduce moving picture data, and supplies the reproduced data to the display unit 1202 through the LCD control unit 1232. Thereby, for example, the moving picture data included in the moving picture file that is linked to the home page is displayed. At the same time, the audio processing unit 1235 converts the audio data into an analog audio signal, and then supplies the analog audio signal to the voice output unit 1208. Thereby, for example, the audio data included in the moving picture file that is linked to the home page is reproduced. Here, a system to which the moving picture coding method and the moving picture decoding method according to any of the aforementioned embodiments is applicable is not restricted to the above-mentioned contents provision system. Recently, digital broadcasting using satellites or terrestrial waves is talked frequently, and the image coding apparatus and the image decoding apparatus according to the above embodiments is applicable also to a digital broadcasting system as shown in Figure 42. More specifically, a code bit stream corresponding to video information is transmitted from a broadcast station 1409 to a satellite 1410 such as a communication satellite or a broadcast satellite, via radio communication. When the broadcast satellite 1410 receives the coded bit stream corresponding to the video information, the satellite 1410 outputs broadcasting waves, and these waves are "received by an antenna 1406 at home including satellite broadcast receiving facility. For example, an apparatus such as a television (receiver) 1401 or a set top box (STB) 1407 decodes the coded bit stream, and reproduces the video information. Further, the image decoding apparatus according to any of the aforementioned embodiments can be mounted also on a reproduction apparatus 1403 that can read and decode the coded bit stream recorded on a storage medium 1402 such as a CD or a DVD (recording medium). In this case, a reproduced video signal is displayed on a monitor 1404. The moving picture decoding apparatus may be mounted on the set top box 1407 that is connected to a cable for cable television 1405 or an antenna for satellite/terrestrial broadcast 1406, to reproduce an output of the moving picture decoding apparatus to be displayed on a monitor 1408 of the television. In this case, the moving picture decoding apparatus may be incorporated not in the set top box but in the television. A vehicle 1412 having an antenna 1411 can receive a signal from the satellite 1410 or the base station 1107, and reproduce a moving picture to display the same on a display device of a car navigation system 1413 or the like which is mounted on the vehicle 1412. Further, it is also possible that an image signal can be coded by the moving picture coding apparatus according to any of the aforementioned embodiments and recorded in a recording medium. A specific example of a recording device is a recorder 1420 such as a DVD recorder that records image signals on a DVD disk 1421, and a disk recorder that records image signals on a hard disk. The image signals may be recorded on a SD card 1422. Further, when the recorder 1420 includes the moving picture decoding apparatus according to any of the aforementioned embodiments, the image signals which are recorded on the DVD disk 1421 or the SD card 1422 can be reproduced by the recorder 1420 and displayed on the monitor 1408. Here, the structure of the car navigation system 1413 may include, for example, the components of the portable telephone shown in figure 41 other than the camera unit 1203, the camera interface unit 1233 and the image coding unit 1242, and the same apply to the computer 1111, or the television (receiver) 1401. Further, as the terminal such as the portable telephone 1114, one of three types of terminals: a transmission-receiving type terminal having both of an encoder and a decoder, a transmission terminal having only an encoder, and a receiving terminal having only a decoder can be mounted. As described above, the moving picture coding method or the moving picture decoding method according to any of the aforementioned embodiments is applicable to any of the above- mentioned devices or systems, whereby the effects as described in the above embodiments can be obtained. Moreover, it is needless to say that the embodiments of the present invention and the applications thereof are not restricted to those described in this specification. APPLICABILITY IN INDUSTRY As described above, in the moving picture coding method and the moving picture decoding method according to the present invention, when a target picture to be coded or decoded is a B picture, a forward picture that is positioned closest to the target picture can be used as a reference picture for the target picture, whereby prediction accuracy in motion compensation for the B picture is increased, resulting in enhanced coding efficiency. Particularly, these methods are useful in data processing for transferring or recording moving picture data. We claim: 1. A moving picture decoding method for decoding a bit stream corresponding to plural pictures constituting a moving picture, comprising: a decoding step of, when decoding a target picture to be decoded, decoding each block in the target picture in either I mode decoding in which a block in the target picture is decoded without referring to another picture, or P mode decoding in which a block in the target picture is predictively decoded with reference to one already-decoded picture, wherein the target picture which has been decoded is stored in a reference picture memory as a candidate for a reference picture, on the basis of candidate picture information which is included in the bit stream and indicates whether or not the target picture is the candidate for the reference picture when decoding another picture that follows the target picture, and in the P mode decoding in which a block in the target picture is predictively decoded, one picture is determined on the basis of reference picture information included in the bit stream, among already-decoded pictures which are stored in the reference picture memory as candidates for the reference picture, and the block is predictively decoded with reference to one determined picture. 2. A moving picture decoding method for decoding a bit stream corresponding to plural pictures constituted a moving picture, comprising: a decoding step of, when decoding a target picture to be decoded, decoding each block in the target picture in one decoding mode among I mode decoding in which a block in the target picture is decoded without referring to another picture, P mode decoding in which a block in the target picture is predictively decoded with reference to one already-decoded picture, and B mode decoding in which a block in the target picture is predictively decoded with reference to two already-decoded pictures, wherein the target picture which has been decoded is stored in a reference picture memory as a candidate for a reference picture, on the basis of candidate picture information which is included in the bit stream and indicates whether or not the target picture is the candidate for the reference picture when decoding another picture that follows the target picture, and in the P mode decoding in which a block in the target picture is predictively decoded, one picture is determined on the basis of reference picture information included in the bit stream, among already-decoded pictures which are stored in the reference picture memory as candidates for the reference picture, and the block is predictively decoded with reference to one determined picture, and in the B mode decoding in which a block in the target picture is predictively decoded, two pictures are determined on the basis of the reference picture information included in the bit stream, among already decoded pictures which are stored in the reference picture memory as the candidates for the reference pictures, and the block is predictively decoded with reference to two determined pictures. According to the present invention, a moving picture coding apparatus (70) for performing inter-picture predictive coding for pictures constituting a moving picture is provided with a coding unit (103) for performing predictive error coding for image data; a decoding unit (105) for performing predictive error decoding for an output from the coding unit (103); a reference picture memory (117) for holding output data from the decoding unit (105); and a motion vector detection unit (108) for detecting motion vectors on the basis of the decoded image data stored in the memory. When coding a B picture as a target picture, information indicating whether or not the target picture should be used as a reference picture when coding another picture is added as header information. Therefore, in a decoding apparatus for decoding a bit stream Bs outputted from the moving picture coding apparatus (70), management of a memory for holding the reference picture can be facilitated on the basis of the header information.