FORM 2
THE PATENTS ACT, 1970
(39 of 1970)
&
THE PATENTS RULES, 2003
COMPLETE SPECIFICATION
(See Section 10; rule 13)
“FAST CAMERA AUTO-FOCUS”
Sony Corporation of the address: 1-7-1 Konan, Minato-ku, Tokyo, 108-0075,
Japan and Sony Electronics Inc. of the address: 1 Sony Drive, Park Ridge, New
Jersey 07656, United States of America.
The following specification particularly describes and ascertains the nature of this invention and
the manner in which it has to be performed:
2
FAST CAMERA AUTO-FOCUS
RELATED APPLICATIONS
[0001] This patent application is related to the co-pending U.S. Patent Applications, entitled “METHOD
AND APPARATUS FOR GENERATING A DEPTH MAP UTILIZED IN AUTOFOCUSING”, application
no. 11/473,694, and “REDUCED HARDWARE IMPLEMENTATION FOR A TWO-PICTURE DEPTH
MAP ALGORITHM”, application no. 12/111,548. The related co-pending applications are assigned to the
same assignee as the present application.
FIELD OF INVENTION
[0002] This invention relates generally to image acquisition, and more particularly to a fast camera autofocus.
COPYRIGHT NOTICE/PERMISSION
[0003] A portion of the disclosure of this patent document contains material which is subject to copyright
protection. The copyright owner has no objection to the facsimile reproduction by anyone of the patent
document or the patent disclosure as it appears in the Patent and Trademark Office patent file or records, but
otherwise reserves all copyright rights whatsoever. The following notice applies to the software and data as
described below and in the drawings hereto: Copyright © 2007, Sony Electronics, Incorporated, All Rights
Reserved.
BACKGROUND
[0004] Auto-focus in a camera automatically focuses a camera lens on a three-dimensional scene, without
the need for the camera user to manually focus the camera lens. There are many auto-focus schemes known in
the art. A traditional auto-focus scheme searches for the peak of the auto-focus curve (e.g., peak-hold-integrate
method, hill-climbing, etc.). Typically, there is a different auto-focus curve for different camera optical
settings (camera type, lens type, aperture, focal length, shutter speed, film or sensor speed, etc.). In addition,
the auto-focus curve will be a function of the scene itself, and the region analyzed. In one embodiment, a
camera employs a hill climbing scheme that computes the auto-focus curve from image gradients. As the
region of analysis becomes more and more in focus, the image gradients will become larger and larger. The
objective of the auto-focus algorithm is to reach the peak of the auto-focus curve using a minimum number of
camera focus positions. In general, hill-climbing auto-focus schemes require a large number of pictures of the
three-dimensional scene in order to converge to the desired focus position.
3
SUMMARY
[0005] A camera auto-focuses on a three-dimensional scene using computed blur differences between
different images of that scene. The camera computes the blur difference between two images of the scene
acquired at two different focus positions. The camera uses the computed blur difference for a given scene
location to predict a new focus position. This procedure repeats until focus is achieved at the given scene
location.
[0006] The present invention is described in conjunction with systems, clients, servers, methods, and
machine-readable media of varying scope. In addition to the aspects of the present invention described in this
summary, further aspects of the invention will become apparent by reference to the drawings and by reading
the detailed description that follows.
BRIEF DESCRIPTION OF THE DRAWINGS
[0007] The present invention is illustrated by way of example and not limitation in the figures of the
accompanying drawings in which like references indicate similar elements.
[0008] Figure 1 illustrates one embodiment of an imaging system 100.
[0009] Figure 2A illustrates one embodiment of a picture of a three dimensional scene.
[0010] Figure 2B illustrates one embodiment of another picture of the three dimensional scene.
[0011] Figure 3 illustrates blur differences of objects in the two pictures.
[0012] Figure 4 illustrates modeling the blur difference between an object in the two pictures.
[0013] Figure 5A (prior art) is a graph of the auto-focus hill-climbing algorithm.
[0014] Figure 5B is a graph illustrating one embodiment of fast auto-focus algorithm.
[0015] Figure 6 is a flow diagram of a method for performing fast auto-focus.
[0016] Figure 7 is a diagram illustrating one embodiment of focusing on a test image for a picture number
representing the focused image.
[0017] Figure 8 is a diagram illustrating one embodiment of a blur difference between two pictures of the
test image.
[0018] Figure 9 is a diagram illustrating a calculated blur difference between the two pictures.
[0019] Figure 10 is a diagram illustrating one embodiment of focusing on a test image for a picture
number that is out of focus.
[0020] Figure 11 is a diagram illustrating one embodiment of a blur difference between two pictures of
the test image.
[0021] Figure 12 is a diagram illustrating a calculated blur difference between the two pictures as
compared to the calculated blur difference in Figure 9.
[0022] Figure 13 is a diagram illustrating one embodiment of focusing on a test image for a picture
number that is out of focus.
[0023] Figure 14 is a diagram illustrating one embodiment of a blur difference between two pictures of
the test image.
4
[0024] Figure 15 is a diagram illustrating a calculated blur difference between the two pictures as
compared to the calculated blur difference in Figures 9 and 12.
[0025] Figure 16 is a graph illustrating a plot of the blur differences versus reference picture number.
[0026] Figures 17A-B are graphs illustrating shifting plots of blur differences versus reference picture
number.
[0027] Figure 18 is a diagram illustrating one embodiment of a set of blur difference reference curves.
[0028] Figure 19A-D are graphs of one embodiment of method 600 illustrated in Figure 6.
[0029] Figures 20-22 is one embodiment of a set of images illustrating the convergence of the fast autofocus
at the mask location shown.
[0030] Figure 23 illustrates a set of masks used to converge the fast auto-focus results as illustrated in
Figures 24-30.
[0031] Figures 24-30 are graphs illustrating fast auto-focus convergence results for the different image
masks of Figure 23.
[0032] Figure 31 is a diagram illustrating one embodiment of an image device control unit that includes
an auto-focus unit.
[0033] Figure 32A is a diagram of one embodiment of an operating environment suitable for practicing
the present invention.
[0034] Figure 32B a diagram of one embodiment of a computer system suitable for use in the operating
environment of Figure 32A.
[0035] Figure 33 is a block diagram illustrating one embodiment of an imaging system that computes the
blur difference between two pictures of a test image using two sensors at different focusing distances from the
test subject.
DETAILED DESCRIPTION
[0036] In the following detailed description of embodiments of the invention, reference is made to the
accompanying drawings in which like references indicate similar elements, and in which is shown by way of
illustration specific embodiments in which the invention may be practiced. These embodiments are described
in sufficient detail to enable those skilled in the art to practice the invention, and it is to be understood that
other embodiments may be utilized and that logical, mechanical, electrical, functional, and other differences
may be made without departing from the scope of the present invention. The following detailed description is,
therefore, not to be taken in a limiting sense, and the scope of the present invention is defined only by the
appended claims.
[0037] As described above, a traditional auto-focus scheme searches for the peak of the auto-focus curve
(e.g., peak-hold-integrate method, hill-climbing, etc.). The auto-focus information contained in the curve is a
qualitative measure of image defocus, or blurring, and an auto-focus scheme attempts to minimize image
blurring. Because an auto-focus scheme measures qualitative blurring and blurring is related to depth, the
auto-focus curve contains depth information. This is because the peak of the curve corresponds to the in-focus
5
position. For example, consider the case where the scene consists of a single object on a constant background.
If the depth of the object is also known, the camera could predict a focus position at which the object is in
focus. As described below with reference to Figure 3, because blur differences are related to depth, blur
differences can be used in an auto-focus scheme to determine the picture that is in-focus, or equivalently, the
focus position for the three dimensional scene.
[0038] Figure 1 illustrates one embodiment of camera 100. In Figure 1, camera 100 comprises lens 102,
sensor 104, control unit 106, storage 108, and lens opening 110. Camera 100 may be digital or film still
camera, video camera, surveillance camera, robotic vision sensor, image sensor, etc. Sensor 104 captures an
image of a scene through lens 102. In one embodiment, camera 100 comprises one sensor. Alternatively,
camera 100 can have multiple sensors at different focusing distance from test subject.
[0039] Sensor 104 can acquire a still picture, such as in a digital or film still camera, or acquire a
continuous picture, such as a video or surveillance camera. In addition, sensor 104 can acquire the image
based on different color models used in the art, such as Red-Green-Blue (RGB), Cyan, Magenta, Yellow,
Green (CMYG), etc. Control unit 106 typically manages the sensor 104 automatically and/or by operator
input. Control unit 106 configures operating parameters of the sensor 104 and lens 102 such as, but not limited
to, the lens focal length, fl, the aperture of the lens, A, lens focus position, and (in still cameras) the lens shutter
speed. In addition, control unit 106 may incorporate an auto-focus unit 120 (shown in phantom) that controls
the automatic focusing of lens 102 on the scene. Embodiments of this invention are incorporated into autofocus
unit 120. The image(s) acquired by sensor 104 are stored in the image storage 108.
[0040] Figure 2A illustrates one embodiment of a picture of a three dimensional scene. In Figure 2,
camera 202 prepares to take a picture by focusing on three-dimensional scene 200. In one embodiment,
camera 202 is camera 100 as illustrated in Figure 1. Three-dimensional scene 200 comprises car 204, person
206, mountain backdrop 208, and sun 210. Each of the objects in three-dimensional scene 200 is at different
distances from camera 202. For example, based on distance scale 224, car 204 is closest to camera 202,
followed in order by person 206, mountain backdrop 208, and sun 210. With camera focusing on car 204 at
distance 222, objects at distances other than distance 222, specifically, greater than or equal to 1 DOF (depth
of field) from the distance 222, will not be in focus. For example, at certain camera settings with camera 202
focusing on car 204, the image resulting from this setting would result with a blurred person 206. Similarly,
mountain backdrop 208 and sun 210 would be blurry as well because these objects are farther still from car
204.
[0041] Picture 214 illustrates this result. In picture 214, car 216 is in focus and not blurry. Blurriness in
picture 214 is represented with a thicker border for each object in the picture. In this example, car 216 has a
relatively thin border. Person 218 has a thicker border. Mountain backdrop 220 has a thicker border 226 and
sun 222 has the thickest border, 228. This represents that sun 222 is blurrier than mountain backdrop 220,
which is blurrier than person 218, which in turn is blurrier than car 216.
6
[0042] Figure 2B illustrates one embodiment of another picture 254 of the three dimensional scene 200
with a different focus position of camera 202. In Figure 2B, camera 202 focus position 252 is before car 204,
specifically, greater than or equal to 1 DOF (depth of field) from distance 222. Thus, the image of car 204 in
picture 254 is blurrier than in the picture 214. As in picture 214, picture 254 comprises person 258, mountain
backdrop 260, and sun 262. While in picture 254, car 256 is blurry with a slightly thicker border as compared
with the focused car 216 in picture 214. Thus, there is a blur difference between car 216 in picture 214 and car
256 in picture 254 based on the different focus positions of camera 202. Person 258, mountain backdrop 260
(with border 261), and sun 262 (with border 263) also have a blur difference. Each of the objects in picture
254 has thicker border than those in picture 214, indicating that these objects are more blurred than those in
picture 214.
[0043] Figure 3 illustrates the absolute blur for different objects in Figures 2A and 2B using pillboxes of
different shapes. The blur difference between car 216 in picture 214 and car 256 in picture 254 can be seen by
examining the difference heights and widths of the two pillboxes shown in 302A-B. This blur difference
results because two different focus positions were used to capture the pictures of Figure 2A and 2B. A pillbox
is mathematical mask used to represent the blurring of an object in an image. As is known in the art, a pillbox
is a blur convolution consisting of equal valued pixels in a circular area centered on a reference pixel. A tall
thin pillbox represents an object that has undergone less blurring versus a shorter, squatter pillbox. In this
Figure, pillbox 302A is taller and thinner than pillbox 302B. Because pillbox 302A is taller than pillbox 302B,
there is a blur difference between the two pillboxes, represented by delta blur 310. Delta blur 310 is used for
fast auto focus, and to obtain distance information for the object.
[0044] In addition, Figure 3 also illustrates blur differences for the other objects in picture 214 and picture
254. In Figure 3, the blur difference 312 (labeled “delta blur2”) is the difference between blurs 304A-B for
person 218 in picture 214 and person 258 in picture 254, respectively. Blur difference 314 (labeled “delta
blur3”) is the difference between blurs 306A-B for mountain backdrop 220 in picture 214 and mountain
backdrop 260 in picture 254, respectively. Blur difference 316 (labeled “delta blur4”) is the difference
between blurs 308A-B for sun 222 in picture 214 and sun 262 in picture 254, respectively.
[0045] As illustrated in Figure 3, because of the different focus positions, a blur difference exists for the
different objects in the scene. This quantity will be referred to herein as the blur difference. The amount of
blur difference depends on the two focus positions. This blur difference is related to the depth of the object
(Equation (1)):
( , , 1, 2, )
( , , 1, 2, )
( , , 1, 2, )
( , , 1, 2, )
sun number sun
mountain number mountain
person number person
car number car
Depth f fl F D D blur
Depth f fl F D D blur
Depth f fl F D D blur
Depth f fl F D D blur
= Δ
= Δ
= Δ
= Δ
(1)
where f is the function that relates blur difference to depth, fl is the focal length of the lens, Fnumber is the fnumber
of the lens, D1 and D2 are the distances between the lens and the image sensor for two different focus
7
positions, and Δblur is the computed blur difference for a specific picture location. As stated previously, the
blur difference results from capturing two pictures using two different focus positions.
[0046] Figure 4 illustrates the modeling and the computation of the blur difference 312 between an object
in the two pictures in Figures 2A and 2B. In Figure 4, blur difference 312 (labeled “delta blur2”) is computed
by applying a 3x3 matrix 402 between blur models 404A-B. Blur models 404A-B can be Gaussian functions,
pillbox functions, or other blur models known in the art. In one embodiment, blur models 404A-B are
Gaussian functions and 3x3 matrix 402 is a blurring matrix that blurs blur model 404A into 404B. 3x3 matrix
402 can be applied to blur model 404A once or many times depending on the amount blurring needed to
convolve blur model 404A into blur model 404B. In one embodiment, 3x3 matrix 402 represents blurring of
one blur unit. 3x3 matrix 402 is also referred to as a blur kernel. In general, the blur kernel is of dimension
nxn, where n is an integer. In one embodiment, the blur kernel is applied via convolution repeatedly on the
sharper image until a new image is produced that closely matches the more blurred image. Qualitatively, this
is illustrated by the smooth, black curve on blur model 404B. This black curve closely approximates the
pillbox 304B after the blur kernel has been applied m times. The number of times the blur kernel has been
applied is also called the iterations number.
[0047] Figure 5A (prior art) is a graph of the auto-focus hill-climbing algorithm. This algorithm finds the
peak of auto-focus curve by calculating gradient information in the given scene. However, this algorithm can
require up to forty or greater pictures in order to converge on a focused image. For example, in Figure 5A,
curve 504 represents the auto-focus curve for a location in a scene. Point 502 represents the picture number
corresponding to the in focus position of curve 504. The in focus position is identified by the peak of curve
504. Different points 500 on curve 504 are different picture numbers for a given camera setting. In one
embodiment, a picture number is a measure of the focus distance, or focus position, of the lens on camera 100.
Picture numbers can be a numbered sequence representing different focus distances. For example, a sequence
of picture numbers for a particular lens combination can be focusing distance of 35, 40, 60, 70, 80, 100, 125,
150, 200, 300, 500, 1000 centimeters and infinity. In one embodiment, each picture number represents a fixed
depth of field difference from an adjacent picture number. For example, in this embodiment, picture number n-
1, n, n+1 can be separated by ½, 1, 2, etc. depth of fields. Analogously, each picture number also represents a
fixed depth of focus difference.
[0048] In Figure 5A, the hill-climbing algorithm works by recording images at different pictures numbers
to compute auto-focus information. The objective of these algorithms is to find the maximum gradient value
corresponding to a given location in the scene. The algorithm performs a search until the picture number
corresponding to point 502 is found. While this algorithm eventually determines the optimal point 502 on the
curve, this algorithm tends to go back and forth around point 502. This results in a large number of pictures for
convergence.
[0049] In contrast, using the blur information resulting from different images captured at different picture
numbers can give fast auto-focus convergence. In the event that the blur is spatially varying in the image, we
8
restrict our attention to a particular scene location in the image. Figure 5B is a graph illustrating one
embodiment of fast auto-focus. In Figure 5B, the fast auto-focus predicts a new picture number on auto-focus
curve 504 using blur difference information from other images. In one embodiment, a camera calculates the
blur difference between two images. Using the blur difference, the camera moves the current focus position to
a new focus position. Two additional pictures are then captured. The process repeats, until the computed
change in blur is less than a preset threshold. (In practice, this corresponds to an iteration number that is
smaller than a preset threshold. Recall, the iteration number is the number of times the blur kernel is applied
to the sharper of the two captured images via convolution, in order to produce an image that has a blur
equivalent to the second, more blurred, captured image.) The lens moves to the new and final focus position
and the final picture is captured. In the event that the blur is spatially varying in the image, attention is
restricted to a specific scene location in the image.
[0050] In practice, since it is not known a priori which image contains more blur, the process is applied in
both directions. This is described in detail in the patent “REDUCED HARDWARE IMPLEMENTATION
FOR A TWO-PICTURE DEPTH MAP ALGORITHM”, application no. 12/111,548.
[0051] Figure 6 is a flow diagram of a method 600 for performing fast auto-focus. In one embodiment,
auto-focus unit 120 executes method 600 to assist in auto-focusing camera lens 102 on the scene. In Figure 6,
at block 602, method 600 determines a starting reference picture. Method 600 can select a reference picture
number at infinite focus, a picture number at the lens minimum focus distance, or a picture number somewhere
in between.
[0052] At block 603, method 600 captures two pictures. If the reference picture number is n, the first
picture is captured at n and the second picture is captured at n + shift, where shift is an integer value. As
stated previously, picture number n and n + shift can be separated by ½, 1, 2, etc. depth of fields.
Analogously, each picture number pair also represents a fixed depth of focus difference.
[0053] At block 604, method 600 computes the blur difference between the pictures captured at n (the
reference picture number) and n + shift (the second captured picture). In one embodiment, method 600
calculates the blur difference between the two images by convolving one image into another image using a
convolving blur kernel. In this embodiment, method 600 applies a blur kernel one or multiple times to
determine the blur difference between to the whole or parts of two images. In another embodiment, method
600 applies a blur kernel to part of one image by using a mask. A mask restricts the evaluation and/or
processing of an image to part of the image defined by the mask. Masks are further described with reference to
Figures 20-23 below. In addition, method 600 can compute the blur difference using full or reduced resolution
images. Computing a blur difference is further described in the co-pending U.S. Patent Applications, entitled
“METHOD AND APPARATUS FOR GENERATING A DEPTH MAP UTILIZED IN AUTOFOCUSING”,
Application no. 11/473,694 and “REDUCED HARDWARE IMPLEMENTATION FOR A TWO-PICTURE
DEPTH MAP ALGORITHM”, application no. 12/111,548. In one embodiment, method 600 determines the
blur difference in the number of iterations needed to convolve one image to another using a 3x3 matrix (blur
9
kernel) that represents one blur unit. In general, the blur kernel can be nxn. Thus, blur differences can be
measured in iterations.
[0054] At block 605, method 600 compares the computed blur difference values and selects the ideal
value. This is described in the patent “REDUCED HARDWARE IMPLEMENTATION FOR A TWOPICTURE
DEPTH MAP ALGORITHM”, application no. 12/111,548.
[0055] At block 606, the appropriate blur difference reference curve is selected from a set of blur
difference reference curves. In one embodiment, the set of curves will be identical. That is, each curve will be
a shifted version of a reference curve. In another embodiment, a set of curves with different characteristics
will exist. The appropriate blur difference reference curve is determined by finding the curve whose zero
crossing location is closest to the current reference picture number n. The generation of such a set of blur
difference reference curves is given in Figures 7-17. One embodiment that uses the computed blur difference
is further described with reference to Figures 19-22 below.
[0056] At block 607, method 600 computes the new reference picture number. In order to determine this
quantity, the appropriate blur difference reference curve needs to be selected from a set of blur difference
reference curves. The computed blur difference quantity is applied to the vertical axis of the selected curve to
determine the new picture number located on the horizontal axis.
[0057] At block 608, method 600 instructs the camera to move the lens to a new reference picture number
ni. The new picture number results from blocks 604 to 607.
[0058] At block 610, method 600 determines if the blur difference/iterations number is less than a predetermined
threshold. In one embodiment, method 600 uses the absolute value of the blur difference/iterations
quantity. If so, the image is focused and method 600 instructs the camera to capture the final picture at picture
number location ni using block 612.
[0059] If the blur difference/iterations number is not less than the pre-determined threshold, method 600
instructs the camera to capture two pictures at block 614. The first picture is captured at picture number ni
while the second picture is captured at picture number ni + shift. Execution proceeds to block 604 above.
[0060] As discussed above, method 600 uses the computed blur difference from two captured pictures to
predict a new reference picture number. Figures 7-15 illustrate one embodiment of the results of method 600
for computing different blur differences at different picture numbers. Figures 7-9 are diagrams illustrating the
blur difference between two blurred step edge images. Figure 7 is a diagram illustrating one embodiment of
focusing on a test subject 710 for a picture number representing the focused image. In Figure 7, camera 702
focuses on test subject 710. Camera 702 can focus on test subject 710 at one of the many different picture
numbers 704 along a distance 706. In this embodiment, camera 702 can focus on test subject 710 at the
minimal picture number 712 at 35 cm (or closer) or at the picture number at infinity 714. The optimally
focused image is at picture number 708. The minimal picture number depends on the type of lens 102 camera
100 uses.
10
[0061] Figure 8 is a diagram 800 illustrating one embodiment of a blur difference between two pictures of
the test image. In Figure 8, camera 802 focuses on test subject 710 at picture number 808. As in Figure 7,
camera 802 can focus on test subject 710 at a different picture numbers 804 along a distance 806. For
reference, diagram 700 of camera 702 focusing on test image 710 at picture number 708 is superimposed on
Figure 8.
[0062] Figure 9 is a diagram 900 of one embodiment illustrating a calculated blur difference between two
images taken at picture numbers 708 and 808. In Figure 9, graph 902 illustrates the blur difference between
images taken at picture number 708 and picture number at 808. Point 908 on graph 902 represents this blur
difference and is plotted on graph 902 as the number of iterations used to convolve the image at picture
number 808 to the image at picture number 708. In this embodiment, point 908 is close to zero iterations.
Images at picture numbers close to or at the optimal focusing location have smaller blur differences than those
for images at adjacent picture numbers for identical shift values. In this embodiment, small blur differences
correspond to small iteration numbers because fewer numbers of blur iterations is needed to convolve one
image into another. Thus, the amount of blur can be measured in the number of iterations used to convolve one
image into another. In contrast, images far from the optimal focusing can have large blur differences with
large numbers of iterations.
[0063] Figures 10-12 are diagrams illustrating the blur difference between two blurred step edge images
and comparing this blur difference with the one calculated in Figure 9. Figure 10 is a diagram 1000
illustrating one embodiment of the focusing on test subject 1010 for a picture number that is out of focus. In
Figure 10, camera 1002 focuses on test subject 1010 at picture number 1008. In this embodiment, picture
number 1008 is closer to camera 1002 than the optimally focused picture at picture number 1012. The image
at picture number 1008 is out of focus and blurred as compared with the image at picture number 708 in
Figure 7.
[0064] Figure 11 is a diagram illustrating one embodiment of a blur difference between two pictures of
the test subject 1010. In Figure 11, camera 1102 focuses on test subject 1010 at picture number 1108. As in
Figure 10, camera 1102 can focus on test subject 1010 at different picture numbers 1104 along a distance
1106. For reference, diagram 1000 of camera 1002 focusing on test image 1010 at picture number 1008 is
superimposed on Figure 11.
[0065] Figure 12 is a diagram illustrating a calculated blur difference between the two pictures as
compared to the calculated blur difference in Figure 9. In Figure 12, graph 1202 illustrates the blur difference
between images taken at picture numbers 1008 and 1108. Point 1210 on graph 1202 represents this blur
difference and is plotted on graph 1202 as the number of iterations to convolve image at picture number 1108
to the image at picture number 1008. In addition, the blur difference calculated in Figure 9 is at point 1208.
Point 1210 represents a larger number of iterations as compared with point 1208. Point 1210 is negative
because the picture captured at picture number 1008 is more blurred than the picture captured at picture
number 1108. For reference, the diagrams of Figures 10 and 11 are superimposed on Figure 12.
11
[0066] Figures 13-15 are diagrams illustrating the blur difference between two blurred step edge images
and comparing this blur difference with the one calculated in Figures 9 and 12. Figure 13 is a diagram 1300
illustrating one embodiment of the focusing on test subject 1310 for a picture number that is out of focus. In
Figure 13, camera 1302 focuses on test subject 1310 at picture number 1308. In this embodiment, picture
number 1308 is further from camera 1302 than the optimally focused picture at picture number 1312. The
image at picture number 1308 is out of focus and blurred as compared with the image at picture number 708 in
Figure 7.
[0067] Figure 14 is a diagram illustrating one embodiment of a blur difference between two pictures of
the test subject 1310. In Figure 14, camera 1402 focuses on test subject 1310 at picture number 1408. As in
Figure 10, camera 1402 can focus on test subject 1310 at different picture numbers 1404 along a distance
1406. For reference, diagram 1300 of camera 1302 focusing on test subject 1310 at picture number 1308 is
superimposed on Figure 14.
[0068] Figure 15 is a diagram illustrating a calculated blur difference between the two pictures as
compared to the calculated blur difference in Figures 9 and 12. In Figure 15, graph 1502 illustrates the blur
difference between images taken at picture numbers 708 and 808, 1008 and 1108 and 1308 and 1408. Point
1512 on graph 1502 represents the latter blur difference and is plotted on graph 1502 as the number of
iterations to convolve image at picture number 1308 to the image at picture number 1408. In addition, the blur
difference calculated in Figures 9 and 12 are at points 1508 and 1510, respectively. Point 1512 represents a
larger number of iterations as compared with point 1508. In contrast with point 1510, point 1512 is positive
because the picture captured at picture number 1308 is less blurred than the picture captured at picture number
1408. For reference, the block diagrams of Figures 13 and 14 are superimposed on Figure 15.
[0069] Figure 16 is a graph 1602 illustrating a plot of the blur iterations versus reference picture number.
In Figure 16, points 1508, 1510 and 1512 represent the calculated blur differences in Figures 9, 12, and 15
along iterations axis 1604 and picture number axis 1606. Curve 1608 can be drawn through points 1508, 1510,
and 1512. As illustrated in Figure 16, curve 1608 is nearly linear in between points 1508, 1510, and 1512.
Furthermore, curve 1608 transitions from negative to positive at point 1610. Zero iteration denotes the focused
point for test subject 1310. In one embodiment, a zero iterations value can result if the two captured pictures
have the same amount of blur. In another embodiment, the iteration number will change from negative to
positive for each successive two picture capture. The picture number associated with this negative to positive
transition closely approximates the focused point. Thus, blur difference can be used to predict where an image
is focused by developing curve 1608 for an image and predicting which picture number has zero iterations on
curve 1608, or, in the typical case, which picture number is associated with the negative to positive transition
in iterations value. In one embodiment, the use of this curve for prediction can be employed in method 600 in
Figure 6 by auto-focus unit 120 above.
[0070] In one embodiment, a blur difference curve, such as curve 1608 in Figure 16, can be used to
determine the optimally focused picture number for a given three dimensional scene and camera optical
12
settings. In another embodiment, the blur difference can be modeled using polynomial, interpolation, or other
algorithms known in the art to approximate a curve. Furthermore, in these embodiments, a curve can be
constructed from computed blur differences to predict an optimal focus distance.
[0071] In alternate embodiments, multiple curves can be used to determine the optimally focused picture
number. In one of these embodiments, this is used when the slope of the curve is not the same for test subjects
1706 and 1756. Figures 17A-B are graphs illustrating shifting plots of blur iterations versus reference picture
number. In Figure 17, curves 1702 and 1752 are blur difference curves for test subject 1706 and 1756 at
different picture numbers. In Figure 17A, curve 1702 transitions from negative to positive at point 1704 while
curve 1752 transitions from negative to positive at point 1754. As illustrated in Figures 17A-B, different blur
difference curves can result when test subject 1706/1756 changes. Thus, an auto-focus algorithm uses different
blur difference reference curves for focusing and to determine a focus distance.
[0072] Figure 18 is a block diagram illustrating one embodiment of a set of empirical blur difference
reference curves. In Figure 18, camera 1802 stores several blur difference curves 1810. There can be blur
difference reference curves for each particular camera optical setting, such a combination of lens type,
aperture, focal distance, focus positions, etc. Alternatively, there can be a set of blur difference reference
curves for many or all possible optical settings.
[0073] In the embodiment illustrated in Figure 18, there is a different reference curve at different picture
numbers for camera 1802. As described above, each picture number represents a different focus distance. Each
curve is plotted as the number of iterations versus reference picture number. Note that each curve goes from
negative iterations to positive number of iterations. Thus, each curve has a picture number where there are
zero iterations. This point defines the focused image for that focus distance. For example, blur difference
reference curve 1816 represents that blur difference curve for a focus distance 1806 of 40 cm. As another
example, blur difference reference curve 1814 represents a focus distance 1808 of 1000 cm.
[0074] In one embodiment, for a given camera, the different reference curves are generated using a
number of test targets. Each camera can have different number of reference curves for different camera
configurations (e.g., the camera’s lens configuration, and other parameters (focal length, aperture, etc)). For
example, one camera at maximum focal length would have four reference curves. At minimum focal length,
this camera would have eleven reference curves. Different cameras could have different numbers of reference
curves.
[0075] Figure 19 A-D are graphs representing different blur difference curves used to predict reference
picture numbers as described in Figure 6, at block 606. Figure 19A is a graph one embodiment of illustrating
the predicting of a next reference picture number using a pair of images. In Figure 19A, camera 1902 acquires
two images 1916A-B at and near distance 1908 (previously defined as n and n + shift) of subject 1904
measured on distance 1906. Because distance 1908 is too close to the camera 1902 as compared with subject
1904, images 1916A-B are blurry. Based on distance 1908, camera 1902 uses the blur difference reference
curve 1912 to predict the next picture number for the auto-focus search. The computed blur difference
13
between images 1916A-B corresponds to the iterations number shown at point 1914. As a result, the lens focus
position is moved from distance 1908 to distance 1932, Figure 19B.
[0076] Figure 19B is a graph illustrating one embodiment of the selection of the next reference picture
number based on two images acquired at and near distance 1932. In Figure 19B, camera 1902 acquires images
1930A-B at and near distance 1932 (previously defined as n1 and n1 + shift, i = 1). Images 1930A-B are still
blurry, but less blurry than images 1916A-B. Based on picture number n1 and n1+ shift at and near distance
1932, camera 1902 selects blur difference curve 1926 to predict the next picture number. The computed blur
difference between images 1930A-B correspond to the iterations number shown at point 1928. As a result, the
lens focus position is moved from distance 1932 to distance 1944, in Figure 19C. For reference, the graph
from Figure 19A is superimposed on Figure 19B.
[0077] Figure 19C is a graph illustrating one embodiment of the selection of the next reference picture
number based on two images acquired at and near distance 1944. In Figure 19C, camera 1902 acquires images
1950A-B at and near distance 1944 (defined as n2 and n2 + shift). Images 1950A-B are less blurry that
images 1930A-B, but have not converged on an in-focus picture. Camera 1902 selects a new picture number
based on curve 1946. The computed blur difference between images 1950A-B correspond to the iterations
number shown at point 1948. Since the iterations number is smaller than a pre-defined threshold, the lens
focus position is moved from distance 1944 to the final distance 1964, in Figure 19D. For reference, the
graphs from Figure 19A-B are superimposed on Figure 19C.
[0078] Figure 19D is a graph illustrating the convergence of the auto-focus search. In Figure 19D, camera
1902 acquires image 1966 at distance 1964. In this embodiment, image 1966 represents the in-focus image of
test subject 1904 that was achieved by acquiring three pairs of image using method 600. For reference, graphs
from Figure 19A-C are superimposed on Figure 19D.
[0079] Figures 20-22 illustrates different examples of a camera using fast auto-focus to converge on a test
subject with different image masks. Figures 20-22 are described in reference to camera 100 in Figure 1 above.
Figure 20 is one embodiment of a set of images illustrating the convergence of the fast auto-focus search using
mask 2002. In Figure 20, mask 2002 is used to converge on a focused image. Mask 2002 restricts the
evaluation area to a subset of the image, such as to one of the faces of the models in the image. In this
embodiment, mask 2004 restricts comparison of blur differences to image area 2004 of mask 2002. The
masks delineate areas in the image containing space invariant blur. For the examples shown, the facial areas
are identified and acquired using face detection technology known in the art. In general a mask of any size
and shape can be applied to an arbitrary location in an image suspected of containing space invariant blur.
[0080] Images 2010A-D illustrate one embodiment of a set of images acquired while converging on rear
model 2008A-D using mask 2002. Mask 2002 blocks off analysis of front model 2006A-D when determining
blur difference using method 600. Image 2010A is at the starting picture number, picture number 209, and
illustrates a blurred back model 2008A with a less blurred front model 2006A. Camera 100 focuses lens 102 to
picture number 133 and acquires image 2010B. Image 2010B has a more in-focus front model 2006B and less
14
in-focus back model 2008B. Camera 100 focuses lens 102 to a new focus position at picture number 112 and
acquires image 2010C. Image 2010C has front model 2006C blurrier than back model 2008C. Back model
2008C is close to being in-focus. Camera 100 focuses lens 102 to in-focus position at picture number 110 and
acquires image 2010D. Image 2010D represents an in-focus image for back model 2008D. Front model 2006D
is out of focus.
[0081] Figure 21 is one embodiment of a set of images illustrating the convergence of the fast auto-focus
search using mask 2102. In Figure 21, mask 2102 is used to converge on a focused image. In this embodiment,
mask 2102 restricts comparison of blur differences to image area 2104 of mask 2102.
[0082] Images 2110A-D illustrate one embodiment of a set of images acquired while converging on rear
model 2108A-D using mask 2102. As in Figure 21, Mask 2102 blocks off analysis of front model 2106A-D
when determining blur difference using for camera 100 using method 600. Camera 100 acquires image 2110A
at the starting picture number 225 and illustrates a more blurred back model 2108A with less blurred front
model 2106A. Camera 100 focuses lens 102 at picture number 113 and acquires image 2110B. Image 2110B
has a less blurred front model 2106B and a more blurred back model 2108B. Camera 100 focuses lens 102 at
picture number 93 and acquires image 2110C. Image 2110C has front model 2106C less blurred than back
model 2108C. Back model 2108 is close to being in focus. Camera 100 focuses lens 102 to in-focus position at
picture number 92 and acquires image 2110D. Image 2110D represents an in-focus image for back model
2108D. Front model 2106D is out of focus.
[0083] Figure 22 is one embodiment of a set of images illustrating the convergence of the fast auto-focus
search using mask 2202. In Figure 22, mask 2202 is used to converge on a focused image. In this embodiment,
mask 2202 restricts comparison of blur differences to image area 2204 of mask 2202.
[0084] Images 2210A-C illustrate one embodiment of a set of images acquired while converging on front
model 2208A-C using mask 2202. Mask 2202 blocks off analysis of rear model 2206A-C when determining
blur difference using for a camera using method 600. Camera 100 acquires image 2210A at the starting
picture number 68. Image 2210A illustrates a blurred front model 2208A with a less blurred back model
2206A. Camera 100 focuses lens 102 at picture number 86 and acquires image 2210B. Image 2210B has an in
focus front model 2206B and less in-focus back model 2208B. Image 2210C represents an in-focus image for
front model 2206C. Back model 2206C is out of focus.
[0085] Figure 23 is a set of masks used in conjunction with fast auto-focus results in Figures 24-30. In
Figure 23, image masks 2302A-G are used to restrict the evaluation and processing area of an image to the
white area of the mask. In this Figure, camera 100 uses image masks 2302A-G to evaluate and/or process
different parts of an image. For example, mask 2302A restricts the processing to the eyes and nose of the
white model. Mask 2302B restricts the processing to the eyes and nose of the black model. Mask 2302C
restricts the processing to the eyes and nose of the white model as in mask 2302A. The distance of the white
model corresponding to mask 2302A is closer than the distance of the white model corresponding to mask
2302C. Mask 2302D restricts the processing to the faces of the front and back models, where the area for the
15
front model is greater than the area of the back model. Mask 2302E is similar to the mask 2302D by restricting
the processing area to the faces of the front and back models, with the area for the front model is greater than
the area of the back model. Mask 2302F has a mask area that includes both the front and back models, where
the area for the front model is greater than the area of the back model. Both encompass a smaller area than
those in 2302E. Mask 2302G has a mask area that includes both the front and back models, with roughly
equal sizes for each model and encompassing a smaller area than mask 2302E or 2302F.
[0086] Figures 24-30 are graphs illustrating fast auto-focus convergence results for the different image
masks of Figure 23. In Figure 24, three sets of graphs illustrate the fast auto-focus using mask 2302A. The
graphs are in-focus graphs 2402A-B that illustrate the focused pictured number based on the starting picture
number, final shift graph 2404A-B that illustrate the shift that occurs right before the lens is moved to the in
focus position, and number of two pictures graphs 2406A-B that illustrates the number of two picture pairs
used to converge on the focused picture.
[0087] In Figure 24, graphs 2402A, 2404A, and 2406A result from camera 100 focusing on the
foreground nose of mask 2302A. Graph 2402A illustrates that the focused picture number varies with the
starting point of the fast auto-focus. For example, the focused picture number varies from 149 to 158
depending on the starting picture number. Graph 2404A reflects the shift that occurs right before the lens is
moved to the in focus position. It is interesting to note that the fast auto-focus converges using 1-5 two picture
pairs as illustrated in graph 2406A.
[0088] In comparison, graphs 2402B, 2404B, and 2406B result from camera 100 focusing on the
background eyes of mask 2302A. Graph 2402B illustrates that the focused picture number varies with the
starting point of the fast auto-focus, but the variance is less than in graph 2402A. For example, the focused
picture number varies from 146 to 149 depending on the starting picture number. Graph 2404B the shift that
occurs right before the lens is moved to the in focus position. It is interesting to note that the fast auto-focus
converges using 1-3 two picture pairs as illustrated in graph 2406B.
[0089] As in Figure 24, Figure 25 illustrates in-focus picture graphs 2502A-B, final shift graphs 2504A-B,
and number of two pictures needed graphs 2506A-B for the foreground nose and background eyes of mask
2302B. For example, graphs 2502A-B illustrate that the in focus picture varies between 156-159 and 151-153,
respectively for the foreground nose and background eyes. Final shift graphs 2504A-B illustrate the shift that
occurs right before the lens is moved to the in focus position, respectively for the foreground nose and
background eyes of mask 2302B. As in Figure 24, the number of two picture pairs needed varies from 1-4 for
the foreground nose and background eyes of mask 2302B.
[0090] Figure 26 illustrates in-focus picture graphs 2602A-B, final shift graphs 2604A-B, and number of
two pictures needed graphs 2606A-B for the foreground nose and background eyes of mask 2302C. For
example, graphs 2602A-B illustrate that the in focus picture varies between 155-159 and 151-154, respectively
for the foreground nose and background eyes. Final shift graphs 2604A-B illustrate the shift that occurs right
before the lens is moved to the in focus position, respectively for the foreground nose and background eyes of
16
mask 2302C. The number of two picture pairs needed varies from 1-5 for the foreground nose and background
eyes of mask 2302C.
[0091] Figure 27 illustrates in-focus picture graphs 2702A-B, final shift graphs 2704A-B, and number of
two pictures needed graphs 2706A-B for the foreground and background faces of mask 2302D. For example,
graphs 2702A-B illustrate that the in focus picture varies between 136-137 and 117-119, respectively for the
foreground and background faces. Final shift graphs 2704A-B illustrate the shift that occurs right before the
lens is moved to the in focus position, respectively for the foreground and background faces of mask 2302D.
The number of two picture pairs needed varies from 1-3 for the foreground nose and background eyes of mask
2302D.
[0092] Figure 28 illustrates in-focus picture graphs 2802A-B, final shift graphs 2804A-B, and number of
two pictures needed graphs 2806A-B for the foreground and background faces of mask 2302E. For example,
graphs 2802A-B illustrate that the in focus picture varies between 128-131 and 108-111, respectively for the
foreground and background faces. Final shift graphs 2804A-B illustrate the shift that occurs right before the
lens is moved to the in focus position, respectively for the foreground and background faces of mask 2302E.
The number of two picture pairs needed varies from 1-3 for the foreground nose and background eyes of mask
2302E.
[0093] Figure 29 illustrates in-focus picture graphs 2902A-B, final shift graphs 2904A-B, and number of
two pictures needed graphs 2906A-B for the foreground and background faces of mask 2302F. For example,
graphs 2902A-B illustrate that the in focus picture varies between 98-100 and 90-93, respectively for the
foreground and background faces. Final shift graphs 2904A-B illustrate the shift that occurs right before the
lens is moved to the in focus position, respectively for the foreground and background faces of mask 2302F.
The number of two picture pairs needed varies from 1-3 for the foreground nose and background eyes of mask
2302F.
[0094] Figure 30 illustrates in-focus picture graphs 3002A-B, final shift graphs 3004A-B, and number of
two pictures needed graphs 3006A-B for the foreground and background faces of mask 2302G. For example,
graphs 3002A-B illustrate that the in focus picture varies between 85-87 and 81-82, respectively for the
foreground and background faces. Final shift graphs 3004A-B illustrate the shift that occurs right before the
lens is moved to the in focus position, respectively for the foreground and background faces of mask 2302G.
The number of two picture pairs needed varies from 1-3 for the foreground nose and background eyes of mask
2302G.
[0095] Figure 31 is a block diagram illustrating one embodiment of an image device control unit that
computes parameters used to auto-focus a lens, such as lens 102 in camera 100 described in Figure 1. In one
embodiment, image control unit 106 contains auto-focus unit 120. Alternatively, image control unit 106 does
not contain auto-focus unit 120, but is coupled to auto-focus unit 120. Auto-focus unit 120 comprises blur
difference curve module 3102, blur difference module 3104, blur comparison module 3106, and picture
number prediction module 3108. Blur difference curve module 3102 selects blur difference reference curve as
17
described in Figure 6, at block 606. Blur difference module 3104 computes the blur difference as described in
Figure 6, at block 604 and Figure 4. Blur comparison module compares the blur difference between two
pictures as described in Figure 6, block 605. Picture number prediction module 3108 predicts the next picture
number as described with reference to Figure 6, block 607.
[0096] Figure 33 is a diagram illustrating one embodiment of an imaging system 3304 that computes the
blur difference between two pictures of a test image using two sensors 3306AB at different focusing distances
from the test subject. In Figure 33, imaging system 3304 comprises lens 3308, splitter 3310, and sensors
3306AB. Sensors 3306AB are at distances 3302AB from lens 3308, respectively. Because sensors 3306AB
have different focusing distances, each sensor acquires a picture with different focal properties. In this
embodiment, imaging system 3304 can acquire images at adjacent pictures number as described with reference
to Figure 8, above. Therefore, imaging system can utilize method 600 as described in Figure 6 above to autofocus
imaging system 3304 with sensors 3306AB. Computing the blur difference for two images acquired by
a two sensor imaging device is further described in “Method of and apparatus for simultaneously capturing
and generating multiple blurred images”, U.S. Patent Publication No. 2007/0189750 and “REDUCED
HARDWARE IMPLEMENTATION FOR A TWO-PICTURE DEPTH MAP ALGORITHM”, application no.
12/111,548.
[0097] In one embodiment, as shown in Figure 32A, a server computer 3201 is coupled to, and provides
data through, the Internet 3205. A client computer 3203 is coupled to the Internet 3205 through an ISP
(Internet Service Provider) 3205 and executes a conventional Internet browsing application to exchange data
with the server 3201. In one embodiment, client 3203 and/or server 3205 can control the auto-focusing of a
coupled camera and/or lens using the fast auto-focus method described in Figures 6 or 18. Optionally, the
server 3201 can be part of an ISP which provides access to the Internet for client systems. The term "Internet"
as used herein refers to a network of networks which uses certain protocols, such as the TCP/IP protocol, and
possibly other protocols such as the hypertext transfer protocol (HTTP) for hypertext markup language
(HTML) documents that make up the World Wide Web (web). The physical connections of the Internet and
the protocols and communication procedures of the Internet are well known to those of skill in the art. Access
to the Internet allows users of client computer systems to exchange information, receive and send e-mails,
view documents, such as documents which have been prepared in the HTML format, and receive content. It is
readily apparent that the present invention is not limited to Internet access and Internet web-based sites;
directly coupled and private networks are also contemplated.
[0098] One embodiment of a computer system suitable for use as server 3201 is illustrated in Figure 32B.
The computer system 3210, includes a processor 3220, memory 3225 and input/output capability 3230
coupled to a system bus 3235. The memory 3225 is configured to store instructions which, when executed by
the processor 3220, perform the methods described herein. The memory 3225 may also store data for a fast
auto-focus mechanism. Input/output 3230 provides for the delivery and display of the data for a fast autofocus
mechanism or portions or representations thereof, and also the input of data of various types for storage,
18
processing or display. Input/output 3230 also encompasses various types of computer-readable media,
including any type of storage device that is accessible by the processor 3220. One of skill in the art will
immediately recognize that the server 3201 is controlled by operating system software executing in memory
3225. Input/output 3230 and related media store the machine-executable instructions for the operating system
and methods of the present invention as well as the data for a fast auto-focus mechanism.
[0099] The description of Figures 32A-B is intended to provide an overview of computer hardware and
other operating components suitable for implementing the invention, but is not intended to limit the applicable
environments. It will be appreciated that the computer system 3240 is one example of many possible computer
systems which have different architectures. A typical computer system will usually include at least a
processor, memory, and a bus coupling the memory to the processor. One of skill in the art will immediately
appreciate that the invention can be practiced with other computer system configurations, including
multiprocessor systems, minicomputers, mainframe computers, and the like. The invention can also be
practiced in distributed computing environments where tasks are performed by remote processing devices that
are linked through a communications network.
[00100] In the foregoing specification, the invention has been described with reference to specific
exemplary embodiments thereof. It will be evident that various modifications may be made thereto without
departing from the broader spirit and scope of the invention as set forth in the following claims. The
specification and drawings are, accordingly, to be regarded in an illustrative sense rather than a restrictive
sense.
19
CLAIMS
What is claimed is:
1. A computerized method comprising:
computing a blur difference between a first and a second image of a three-dimensional scene, the first
image acquired through a camera lens focused at a first picture number and the second image acquired through
the camera lens focused at a second picture number, wherein the first and second picture numbers represent
different focus distances of the camera lens; and
predicting a third picture number based on the computer blur difference, wherein the third picture
number is subsequently used to auto-focus the camera lens on the scene.
2. The computerized method of claim 1, wherein the predicting is further based on a blur difference curve
associated with the first picture number.
3. The computerized method of claim 2, further comprising:
retrieving the blur difference curve from one of a plurality of stored blur difference curves.
4. The computerized method of claim 3, wherein each of the plurality of stored blur difference curves is
associated with a different picture number.
5. The computerized method of claim 2, further comprising:
determining the blur difference curve using the computed blur difference.
6. The computerized method of claim 1, wherein the computing a blur difference is further based on using an
image mask.
7. The computerized method of claim 1, wherein the computing a blur difference comprises:
convolving the first image into the second image with a blur kernel.
8. The computerized method of claim 1, wherein the first picture number is a picture number at one of a
minimal focus distance and an infinite focus distance.
9. The computerized method of claim 1, wherein the first picture number is predicted from a blur difference of
a third and fourth image.
10. The computerized method of claim 1, wherein the second picture number is adjacent to the first picture
number.
11. A machine readable medium having executable instructions to cause a processor to perform a method
comprising:
computing a blur difference between a first and a second image of a three-dimensional scene, the first
image acquired through a camera lens focused at a first picture number and the second image acquired through
the camera lens focused at a second picture number, wherein the first and second picture numbers represent
different focus distances of the camera lens; and
predicting a third picture number based on the computer blur difference, wherein the third picture
number is subsequently used to auto-focus the camera lens on the scene.
20
12. The machine readable medium of claim 11, wherein the predicting is further based on a blur difference
curve associated with the first picture number.
13. The machine readable medium of claim 12, further comprising:
retrieving the blur difference curve from one of a plurality of stored blur difference curves.
14. The machine readable medium of claim 13, wherein each of the plurality of stored blur difference curves is
associated with a different picture number.
15. The machine readable medium of claim 12, further comprising:
determining the blur difference curve using the computed blur difference.
16. The machine readable medium of claim 11, wherein the computing a blur difference comprises:
convolving the first image into the second image with a blur kernel.
17. An apparatus comprising:
means for computing a blur difference between a first and a second image of a three-dimensional
scene, the first image acquired through a camera lens focused at a first picture number and the second image
acquired through the camera lens focused at a second picture number, wherein the first and second picture
numbers represent different focus distances of the camera lens; and
means for predicting a third picture number based on the computer blur difference, wherein the third
picture number is subsequently used to auto-focus the camera lens on the scene.
18. The apparatus of claim 17, wherein the means for predicting is further based on a blur difference curve
associated with the first picture number.
19. The apparatus of claim 18, further comprising:
means for retrieving the blur difference curve from one of a plurality of stored blur difference curves.
20. The apparatus of claim 18, further comprising:
means for determining the blur difference curve using the computed blur difference.
21. The apparatus of claim 17, wherein the means for computing a blur difference comprises:
means for convolving the first image into the second image with a blur kernel.
22. A system comprising:
a processor;
a memory coupled to the processor though a bus; and
a process executed from the memory by the processor to cause the processor to compute a blur
difference between a first and a second image of a three-dimensional scene, the first image acquired through a
camera lens focused at a first picture number and the second image acquired through the camera lens focused
at a second picture number, wherein the first and second picture numbers represent different focus distances of
the camera lens and predict a third picture number based on the computer blur difference, wherein the third
picture number is subsequently used to auto-focus the camera lens on the scene.
21
23. A computerized method substantially as herein described with reference to the foregoing description and
the accompanying drawings.

Dated this 23rd day of September 2009.
To
The Controller of Patents
The Patent Office
India.