DESC:TECHNICAL FIELD
[0001] The present invention relates generally to image sensors and imaging systems. The invention, more particularly, provides a system and method for cross-spectral stereo image registration.

BACKGROUND
[0002] Humans perceive the world in the visible portion of electromagnetic spectrum i.e., to . However, the visible radiations cover only a tiny portion of total electromagnetic spectrum, and there is a lot of information present in other portions of electromagnetic spectrum. With the advancement in imaging sensor technology, it is now possible to sense radiations in specific wavelength ranges (or bands) and generate images of real world scene as observed in those specific wavelength bands.
[0003] Multispectral imaging is a well-known technique to expand knowledge about the surrounding world by capturing images within specific wavelength bands across the electromagnetic spectrum. Each of these wavelength bands provides different characteristics of the scene under observation. These different characteristics are then fused together to complement each other and provide comprehensive understanding of surroundings to the end user. Consequently, multispectral imaging finds its uses in numerous fields including healthcare, weather forecasting, satellite imagery, military applications etc.
[0004] Generally, a multispectral imaging setup includes multiple cameras for capturing images in multiple wavelength bands simultaneously. These cameras are placed in fixed positions and orientations relative to each other. One of the major challenges with this kind of setup is to align or register images captured from multiple cameras in multiple wavelength bands, i.e., the challenge of cross-spectral registration.
[0005] In imaging setups used in applications such as satellite and medical imaging, the registration of images can be done with relative ease using a rigid similarity transformation with some prior information. This is because of the constraints and priors available in these setups, e.g., in satellite imaging the scene under consideration is more or less a flat surface far away from the camera setup and in medical imaging we have the prior knowledge of object being imaged along with fixed (or constrained) distance of object from camera setup.
[0006] However, in unconstrained imaging setup without any prior knowledge of object(s) being imaged (such as those used in robotics, military applications, autonomous vehicles, AI applications, industrial applications etc.), the registration requires a non-rigid transformation, which is generally achieved by using sophisticated stereo matching methods. There is significant amount of prior work on cross-spectral registration for satellite and medical imaging applications, but it is limited for unconstrained imaging applications such as robotics, military applications etc.
[0007] Some recent methods for registration of multispectral satellite images are disclosed in patent documents CN104732532A and CN105631872B. Some recent methods for registration of multispectral medical images are disclosed in patent documents US8855386B2, and CN110544274A. The methods disclosed in the above documents mainly depend on feature matching and are designed to work well when the images under consideration have some common features. However, when the images under consideration are captured in completely non-overlapping wavelength bands, these methods do not perform well because of lack of common features.
[0008] Recent state-of-the-art methods for registering cross-spectral images rely on matching some features which are common across the spectrums. However, when the involved spectrums do not have sufficient overlap, these methods fail due to lack of significant common features.
[0009] There is still a need of an invention which solves the above defined problems and provides a system and a method to register images captured in non-overlapping wavelength bands without the need of feature matching across the wavelength bands.

SUMMARY OF THE INVENTION
[0010] This summary is provided to introduce concepts of the present invention. This summary is neither intended to identify essential features of the present invention nor is it intended for use in determining or limiting the scope of the present invention.
[0011] In accordance with the present invention, a method for cross-spectral stereo image registration is disclosed. The method comprises: capturing one or more images of an object using plurality of cameras; estimating fundamental matrices and rectification transformations for the said cameras; rectifying the images from cameras sensitive to overlapping wavelength bands; estimating stereo correspondence between rectified images from cameras sensitive to overlapping wavelength bands and compute a disparity map; registering images from cameras sensitive to non-overlapping wavelength bands. The step of registering images from cameras sensitive to non-overlapping wavelength bands comprise of: calculating corresponding pixel location between two images from cameras sensitive to overlapping wavelength bands using the calculated stereo correspondence and a disparity map; calculating corresponding pixel location between two images from cameras sensitive to non-overlapping wavelength bands; generating registered image by warping; and interpolating the registered image using corresponding pixel location.
[0012] In one aspect, capturing one or more images using plurality of cameras, comprises: capturing multispectral images using plurality of cameras sensitive to different non-overlapping wavelength ranges.
[0013] In one aspect, capturing one or more images using plurality of cameras comprises: capturing one or more images wherein the captured images are of two different non-overlapping wavelength bands; and capturing an image at a wavelength band having an overlap with either of the two different non- overlapping wavelength bands.
[0014] In another aspect of the invention, a system for non-rigid registration of multispectral images sensitive to different non-overlapping wavelength ranges, is disclosed. The system comprises: a multispectral imaging apparatus comprising plurality of cameras; an image acquisition module for acquiring image data from the cameras; a conventional stereo calibration module configured to estimate fundamental matrices and rectification transformations for the cameras; an image rectification module configured to rectify the images from cameras sensitive to overlapping wavelength bands captured by image acquisition module; a conventional stereo matching module configured to estimate stereo correspondence between rectified images from cameras sensitive to overlapping wavelength bands; a cross-spectral registration module configured to register images from cameras sensitive to non-overlapping wavelength bands. The cross spectral registration module further comprises: a corresponding pixel location calculator module configured to calculate corresponding pixel location between two images from cameras sensitive to overlapping wavelength bands using pre-computed stereo correspondence and disparity map; a cross-spectral pixel correspondence calculator module configured to calculate corresponding pixel location between two images from cameras sensitive to non-overlapping wavelength bands; an image warping and interpolation module configured to generate registered image by warping; and interpolate registered image using corresponding pixel location.
[0015] In one aspect, the multispectral imaging apparatus has three cameras, wherein the first and second cameras are sensitive to two different non-overlapping wavelength bands and the third camera is sensitive to a wavelength band having overlap with one of the first two cameras.
[0016] In one aspect, the corresponding pixel location calculator module is configured to calculate corresponding pixel location between two images from cameras sensitive to overlapping wavelength bands using rectification transformations.
[0017] In one aspect, the corresponding pixel location calculator module is configured to: compute horizontal and vertical epipolar lines from rectified images; and identify intersection of horizontal and vertical epipolar lines.
[0018] In one aspect, a communication module configured to enable communication between the modules; and one or more storage devices configured to store data is disclosed.

BRIEF DESCRIPTION OF ACCOMPANYING DRAWINGS
[0019] The detailed description is described with reference to the accompanying figures. In the figures, the left-most digit(s) of a reference number identifies the figure in which the reference number first appears. The same numbers are used throughout the drawings to reference like features and modules.
[0020] Figure 1A-1C illustrates the physical placement and design of a part of the system for cross-spectral stereo image registration, according to an exemplary implementation of the present invention.
[0021] Figure 2 illustrates an exemplary choice of wavelength bands for the cameras illustrated in Fig. 1, according to an exemplary implementation of the present invention.
[0022] Figure 3 illustrates a conceptual representation of different image planes corresponding to different cameras according to an exemplary implementation of the present invention.
[0023] Figure 4 illustrates a multispectral imaging apparatus, according to an exemplary implementation of the present invention.
[0024] Figure 5 illustrates a flowchart illustrating the general process flow of the method of cross-spectral stereo image registration, according to an exemplary implementation of the present invention.
[0025] Figure 6 illustrates a block diagram of an apparatus for registering an image pair , according to an exemplary implementation of the present invention.
[0026] Figure 7 illustrates a block diagram of the apparatus for estimating stereo correspondence between an image pair , according to an exemplary implementation of the present invention.
[0027] It should be appreciated by those skilled in the art that any block diagrams herein represent conceptual views of illustrative methods embodying the principles of the present invention. Similarly, it will be appreciated that any flow charts, flow diagrams, and the like represent various processes which may be substantially represented in computer readable medium and so executed by a computer or processor, whether or not such computer or processor is explicitly shown.
DETAILED DESCRIPTION
[0028] The various embodiments of the present invention describe about system and method for cross-spectral stereo image registration.
[0029] In the following description, for purpose of explanation, specific details are set forth in order to provide an understanding of the present invention. It will be apparent, however, to one skilled in the art that the present invention may be practiced without these details. One skilled in the art will recognize that embodiments of the present invention, some of which are described below, may be incorporated into a number of systems.
[0030] However, the systems and methods are not limited to the specific embodiments described herein. Further, structures and devices shown in the figures are illustrative of exemplary embodiments of the presently invention and are meant to avoid obscuring of the present invention.
[0031] It should be noted that the description merely illustrates the principles of the present invention. It will thus be appreciated that those skilled in the art will be able to devise various arrangements that, although not explicitly described herein, embody the principles of the present invention. Furthermore, all examples recited herein are principally intended expressly to be only for explanatory purposes to help the reader in understanding the principles of the invention and the concepts contributed by the inventor to furthering the art and are to be construed as being without limitation to such specifically recited examples and conditions. Moreover, all statements herein reciting principles, aspects, and embodiments of the invention, as well as specific examples thereof, are intended to encompass equivalents thereof.
[0032] In one of the embodiments, a system and method for cross-spectral stereo image registration without the need of cross-spectral feature matching is disclosed. To register images from two stereo cameras operating on non-overlapping wavelength ranges, the system uses a third camera whose operating wavelength range has some overlap with any one of first two cameras. The system has placement and alignment of these cameras. The method exploits the alignment of cameras for cross-spectral stereo image registration without the need of cross-spectral feature matching.
[0033] In an embodiment, a system and method of the present invention discloses enabling non-rigid registration (or alignment) of multispectral images captured from two cameras sensitive to different non-overlapping wavelength ranges (or bands) across the electromagnetic spectrum. The system comprises a multispectral imaging apparatus having three cameras (or sensors), of which, first two are sensitive to two different non-overlapping wavelength bands and third one is sensitive to a wavelength band having some overlap with one of the first two cameras. The multispectral imaging apparatus comprises an image acquisition module for collecting/acquiring/capturing image data from the three cameras, a conventional stereo calibration module for estimating fundamental matrices and rectification transformations for the cameras, an image rectification module for rectifying the images from cameras sensitive to overlapping wavelength bands captured by image acquisition module, a conventional stereo matching module for estimating stereo correspondence between rectified images from cameras sensitive to overlapping wavelength bands, a cross-spectral registration module for registering images from cameras sensitive to non-overlapping wavelength bands, a corresponding pixel location calculator module for calculating corresponding pixel location between two images from cameras sensitive to overlapping wavelength bands using pre-computed stereo correspondence and disparity map, a cross-spectral pixel correspondence calculator module for calculating corresponding pixel location between two images from cameras sensitive to non-overlapping wavelength bands, an image warping and interpolation module for generating registered image by warping and interpolating it using corresponding pixel location and a communication module for handling communication between various modules and sub-modules. All user interactions are handled by the multispectral imaging apparatus using a user interface module.
[0034] In an exemplary embodiment, a cross-spectral non-rigid stereo image registration using a third camera is achieved without the need of cross-spectral feature matching.
[0035] In an exemplary embodiment, calculation of corresponding pixel location between two images from cameras sensitive to overlapping wavelength bands is performed using rectification transformations and disparity map.
[0036] In an exemplary embodiment, calculation of corresponding pixel location between two images from cameras sensitive to non-overlapping wavelength bands is performed by computing horizontal and vertical epipolar lines and their intersection.
[0037] In an exemplary embodiment, generation of registered image is performed by warping and interpolating it using corresponding pixel location.
[0038] The present invention provides an imaging system and method for cross-spectral image registration. In particular, the present invention provides a system and method for non-rigid registration (or alignment) of image(s) captured from two cameras sensitive to different non-overlapping wavelength ranges (or bands) across the electromagnetic spectrum. Unlike prior works, which depend on finding common features across different wavelength bands, the system has a third camera sensitive to a wavelength band which overlaps with any of the first two cameras. The system and method disclosed in this invention first registers images from the two cameras whose wavelength bands have some overlap (i.e., the third camera and one of the first two cameras whose wavelength band overlaps with that of the third camera). And then, uses this registration information to register images from first two cameras whose wavelength bands do not have any overlap. The system is capable of registering images captured in different non-overlapping wavelength bands, without the need of matching common features across the wavelength bands.
[0039] The expression "registration" in the instant invention refers to the process of mapping each pixel of an image captured from a camera with a pixel of another image captured from another camera at the same time instant such that the mapped pixels on the two images correspond to same physical point in the real world.
[0040] The expression "cross-spectral registration" in the instant invention refers to a registration process wherein the inputs images have characteristics acquired from two different wavelength bands of electromagnetic spectrum.
[0041] The expression "wavelength band" in the instant disclose refers to a continuous range of wavelengths in the electromagnetic spectrum.
[0042] The expression "camera" in the instant invention refers to any digital device having a hardware and/or software capable of sensing electromagnetic radiations in specific wavelength band and generating image of a real world scene as observed in that wavelength band.
[0043] The expression "image" in the instant invention refers to digital data or digital representation of analog electrical signals coming via analog-to-digital converter from each of the detector or sensor wherein output of individual detectors are multiplexed to the Read-Out Integrated Circuit and represent the brightness and color level measured by the detector and vary from image source to image source having a fixed range depending on the sensor.
[0044] Fig. 1A-1C shows physical placement and alignment of three cameras as disclosed, in front view, side view and top view in the system for cross-spectral stereo image registration. A camera C1 101 and a camera C2 102 represent first two cameras sensitive to two different non-overlapping wavelength bands. In one exemplary embodiment, the camera C1 101 may be normal RGB camera sensitive to wavelength band ranging from to and the camera C2 102 may be an infrared camera sensitive to long-wave-infrared band ranging from to ). A camera C3 103 represents the third camera which is sensitive to a wavelength band having some overlap with any one of the first two cameras. In one exemplary embodiment, the camera C3 103 may be a normal RGB camera sensitive to wavelength band ranging from to or an RGB-IR camera sensitive to wavelength band ranging from to and have overlap with camera C1 101. The placement and orientations of the cameras C1 101, C2 102, and C3 103 are fixed relative to each other as illustrated in Fig. 1. The optical center of each of cameras. Camera C1 101, camera C2 102 and camera C3 103 lie on the same plane. The alignment of camera C1 101 and camera C2 102 is orthogonal to the alignment of camera C3 103 and camera C2 102. The optical axes of camera C1 101 and camera C3 103 may be tilted slightly towards the optical axis of camera C2 102.
[0045] An exemplary choice of wavelength bands for the three cameras is illustrated in Fig. 2. The camera C1 and camera C2 have non-overlapping wavelength bands, while the camera C3 has wavelength band overlapping with that of camera C1.
[0046] Fig.3 illustrates a conceptual representation of different image planes corresponding to different cameras according to the exemplary embodiment of the present invention. Image plane I1 301, Image plane I2 302 and Image plane I3 303 are image plane I1, image plane I2 and image plane I3 corresponding to cameras C1, C2 and C3, respectively. An image plane is a virtual representation of image captured by a camera. Points O, A, B, C and D are real world points such that points A, O, B are collinear and points C, O, D are collinear as shown in Fig. 3. O1, O2 and O3 represent images of point O as captured by camera C1, C2 and C3, respectively. Similarly, A2, B2, C2 and D2 represent images of points A, B, C and D as captured by camera C2. X1X'1, X2X'2 and X3X'3 are the lines of sight of camera C1, C2 and C3 respectively to the point O in real world. From Fig. 3, it can be observed that line of sight of camera C1 passing through points A, O and B is projected as slanted horizontal line on image plane I2. This slanted horizontal line on image plane I2 passes through images of points A, O and B as captured by camera C2 (i.e., through A2, O2 and B2). Similarly, line of sight of camera C3 passing through points C, O and D is projected as slanted vertical line on image plane I2. This slanted vertical line on image plane I2 passes through images of points C, O and D as captured by camera C2 (i.e., through C2, O2 and D2). It can be noted that the two lines on image plane I2 intersects at image O2 of point O. Assuming that for a given image point O1 on image (plane) I1 if one knows the corresponding location of image point O3 in image (plane) I3, one can use the inherent geometrical properties of the imaging apparatus to estimate the two lines on image plane I2. The intersection of these two lines then gives the image point O2 on image (plane) I2 which corresponds to image point O1 on image (plane) I1.
[0047] In a multispectral imaging apparatus where camera C1 and camera C2 do not have any overlap in their wavelength bands, one can choose camera C3 such that its wavelength band has some overlap with camera C1. Because of the overlap in wavelength bands (and hence availability of common features) in images captured from camera C1 and C3, conventional stereo registration methods can estimate pixel wise correspondence between these images. This pixel wise correspondence between images captured by camera C1 and C3 can then be used to compute pixel wise correspondence between images captured by camera C2 and camera C1 (or C3) by the virtue of geometrical properties of imaging apparatus disclosed in this invention. In this way, the present invention registers images captured from cameras having no overlap in their wavelength bands (e.g., camera C2 and C1) without any requirement of cross-spectral feature matching.
[0048] Fig.4 illustrates a schematic block diagram of the multispectral imaging apparatus, according to an exemplary implementation of the present invention. The multispectral imaging apparatus 401 includes of three cameras C1, C2 and C3 sensitive to different wavelength bands placed according to the arrangement disclosed in Fig. 1. An image acquisition module 402 communicates with these cameras and captures image data from the cameras C1, C2 and C3 and stores them in a memory 403. A conventional stereo calibration module 404 performs stereo calibration for camera pairs (C1, C3), (C1, C2) and (C3, C2) to compute respective fundamental matrices, it also computes rectification transformations for camera C1 and C3. An image rectification module 405 reads rectification transformations and performs image rectification on images from camera C1 and C3. A conventional stereo matching module 406 estimates stereo correspondence between rectified images from camera C1 and C3. A cross-spectral image registration module 409 registers the images from camera C1 and C2 (or C3 and C2). A corresponding pixel location calculator module 410 computes pixel wise correspondence between images from camera C1 and C3 using the fundamental matrices and stereo correspondence between those images. A cross- spectral pixel correspondence module 411 computes pixel wise correspondence between images from camera C1 (or C3) and C2. An image warping and interpolation module 412 warps image from camera C2 to align with image from camera C1 (or C3). A user interface 413 allows end user to interact with the system and a communication module 414 handles communication between various modules and sub-modules. Here, the “modules" refers to a functional structure divided according to logic, and the "module" can be implemented by pure hardware, or a combination of software and hardware.
[0049] Figure 5 shows a flow chart for method for cross-spectral stereo image registration. At step 501, images , and are captured simultaneously from cameras C1, C2 and C3, respectively. At step 502, the image pair is rectified, by performing a projective transformation called rectification transformation. The rectification transformation is applied to each of the images and , such that in the resulting image pair, conjugate epipolar lines coincide with aligned image rows. The rectification operation is a conventional step before estimating stereo correspondence and requires the prior knowledge of relative position, orientation, and internal parameters of the two cameras. This prior knowledge is encapsulated in a fundamental matrix, which is a matrix that concisely expresses the relation between two cameras. This fundamental matrix is computed using conventional stereo camera calibration methods and rectification transformations are derived from fundamental matrix. After applying rectification transformations on and , at step 503, a stereo correspondence is estimated between pair of rectified images which thereby generates a disparity map for the image pair . The disparity map is a matrix with number of rows same as number of rows in rectified images and number of columns same as number of columns in rectified images. Each element of the disparity map is a measure of physical distance between camera setup and the real world object corresponding to pixel location . These disparity map values obtained after registering image pair are then used to register image pair (or image pair ) pixel-by-pixel. For every pixel at location in image , at step 504 the corresponding pixel location in image is computed using stereo correspondence obtained in step 503. At step 505, two epipolar lines on image , one for pixel in image and another for pixel in image , are computed and at step 506, intersection of these two epipolar lines on image is computed. The point of intersection of these two epipolar lines on image , say , corresponds to the pixel in image . At step 507, by performing the operations above, for every pixel location in image , a corresponding pixel location in image is estimated, and then the image is warped and interpolated according to this mapping and the warped image is registered with image .
[0050] Figure 6 illustrates a block diagram of the system for registering the image pair . An image acquisition module 601 communicates with cameras C1, C2 and C3 to capture images , and simultaneously. These images are then stored in a volatile memory 602. The fundamental matrix for camera pair (C1,C3) obtained from conventional stereo calibration of the cameras is stored in a non-volatile memory 603 along with rectification transformations and (derived from using conventional techniques) for image and respectively. At module 604, the rectification transformations and are accessed from the non-volatile memory 603 and then applied to image and respectively. The rectified images are sent to module 605 to estimate stereo correspondence between image and , and compute the disparity map . The computed disparity map is then stored in a volatile memory 606.
[0051] Figure 7 illustrates a block diagram of the system for estimating stereo correspondence between the image pair using pre-computed stereo correspondence between the image pair .This requires the prior knowledge of rectification transformations , and two fundamental matrices viz. fundamental matrix for the camera pair (C1,C2) and fundamental matrix for camera pair (C3,C2). The fundamental matrices and rectification transformations are computed using stereo camera calibration methods. For the camera setup illustrated in Fig. 1 of this invention, a matrix is computed considering horizontal stereo alignment between camera C1 and C2, while the matrix is computed considering vertical stereo alignment between camera C3 and C2. For every pixel at location in image , the rectification transformations and for image and respectively are read from a non-volatile memory 701, and corresponding disparity map value is read from a volatile memory 702.The corresponding pixel location in image is computed using Equation 1.

(1)

[0052] The fundamental matrices and for camera pair (C1, C2) and (C3, C2) respectively are read from non-volatile memory 701. These fundamental matrices are used to compute epipolar lines and on image using Equation 2 and 3, respectively.

(2)

(3)
[0053] Since, the matrix is computed considering horizontal stereo alignment between camera C1 and C2, while the matrix is computed considering vertical stereo alignment between camera C3 and C2, the epipolar lines obtained using are horizontal, while those obtained using are vertical. For any given pixel on image , the point of intersection of horizontal and vertical epipolar lines on image gives the pixel location on image corresponding to pixel location on image . 703 computes the intersection of these lines using Equation 4 and stores the computed intersection in volatile memory 704.
,
(4)
[0054] The intersection of epipolar lines on image is computed for each pixel location in image , and then the image is warped according to computed intersection points to get registered image using Equation 5.

(5)

[0055] After warping image , there may be some missing pixels in warped image. These missing pixels are eliminated by interpolating them from their neighbours. Although the invention of Figure 7 shows registering image pair , the same procedure may also be used for registering image pair .
[0056] Thus, the present invention discloses a system and associated method for non-rigid cross-spectral stereo image registration without the need of cross-spectral feature matching. To register images from two stereo cameras operating on non-overlapping wavelength ranges, the system uses a third camera whose operating wavelength range has some overlap with any one of first two cameras. The system discloses a placement and alignment of these cameras and a method that exploits this alignment for cross-spectral stereo image registration without the need of cross-spectral feature matching. The system and method mainly comprises of a multispectral imaging apparatus and an image acquisition module for acquiring multispectral images; a conventional stereo calibration module, an image rectification module, and a conventional stereo matching module for registering image from third camera with image from any one of the first two cameras; a cross-spectral image registration module for registering images from first two cameras. The method first registers image from third camera with image from any one of the first two cameras. This registration information is then used to register images from the first two cameras.
[0057] The person skilled in the art can understand that the whole or part of the steps for achieving the above-described embodiments can be accomplished by hardware, or be accomplished by a program instructing relevant hardware, the program may be stored in a computer readable storage medium, and the above-mentioned storage medium may be a read-only memory, a magnetic disc or an optical disc, FPGA, or using relevant combinations of hardware and software etc.
[0058] The foregoing description of the invention has been set merely to illustrate the invention and is not intended to be limiting. Since modifications of the disclosed embodiments incorporating the substance of the invention may occur to person skilled in the art, the invention should be construed to include everything within the scope of the invention.
,CLAIMS:
1. A method for cross-spectral stereo image registration, comprising:
capturing (501) one or more images of an object using plurality of cameras;
estimating (502) fundamental matrices and rectification transformations for the said cameras;
rectifying the images from cameras sensitive to overlapping wavelength bands;
estimating (503) stereo correspondence between rectified images from cameras sensitive to overlapping wavelength bands and compute a disparity map;
registering images from cameras sensitive to non-overlapping wavelength bands, wherein registering images from cameras sensitive to non-overlapping wavelength bands comprise of:
calculating (504) corresponding pixel location between two images from cameras sensitive to overlapping wavelength bands using the calculated stereo correspondence and a disparity map;
calculating corresponding pixel location between two images from cameras sensitive to non-overlapping wavelength bands;
generating registered image by warping (507); and
interpolating the registered image using corresponding pixel location.

2. The method as claimed in claim 1, wherein capturing one or more images using plurality of cameras, comprises:
capturing multispectral images using plurality of cameras sensitive to different non-overlapping wavelength ranges.

3. The method as claimed in claim 2, wherein the capturing one or more images using plurality of cameras comprises:
capturing one or more images wherein the captured images are of two different non-overlapping wavelength bands; and
capturing an image at a wavelength band having an overlap with either of the two different non- overlapping wavelength bands.
4. The method as claimed in claim 1, wherein calculating corresponding pixel location between two images from cameras sensitive to overlapping wavelength bands is performed using rectification transformations.

5. The method as claimed in claim 1, wherein corresponding pixel location between two images from cameras, comprises:
computing (505) horizontal and vertical epipolar lines from rectified images; and
identifying (506) intersection of horizontal and vertical epipolar lines.

6. A system for non-rigid registration of multispectral images sensitive to different non-overlapping wavelength ranges, the system comprising:
a multispectral imaging apparatus (401) comprising plurality of cameras (C1,C2,C3);
an image acquisition module (402) for acquiring image data from the cameras;
a conventional stereo calibration module (404) configured to estimate fundamental matrices and rectification transformations for the cameras;
an image rectification module (405) configured to rectify the images from cameras sensitive to overlapping wavelength bands captured by image acquisition module;
a conventional stereo matching module (406) configured to estimate stereo correspondence between rectified images from cameras sensitive to overlapping wavelength bands;
a cross-spectral registration module configured to register images from cameras sensitive to non-overlapping wavelength bands, comprising:
a corresponding pixel location calculator module (410) configured to calculate corresponding pixel location between two images from cameras sensitive to overlapping wavelength bands using pre-computed stereo correspondence and disparity map;
a cross-spectral pixel correspondence calculator module (411) configured to calculate corresponding pixel location between two images from cameras sensitive to non-overlapping wavelength bands;
an image warping and interpolation module (412) configured to generate registered image by warping; and interpolate registered image using corresponding pixel location.

7. The system as claimed in claim 6, wherein the multispectral imaging apparatus has three cameras (C1, C2, C3), wherein the first and second cameras (C1, C2) are sensitive to two different non-overlapping wavelength bands and the third camera (C3) is sensitive to a wavelength band having overlap with one of the first two cameras.

8. The system as claimed in claim 6, wherein the corresponding pixel location calculator module is configured to calculate corresponding pixel location between two images from cameras sensitive to overlapping wavelength bands using rectification transformations.

9. The system as claimed in claim 6, wherein corresponding pixel location calculator module is configured to:
compute horizontal and vertical epipolar lines from rectified images; and
identify intersection of horizontal and vertical epipolar lines.

10. The system as claimed in claim 6 the system further comprises:
a communication module (414) configured to enable communication between the modules; and
one or more storage devices (403, 407, 408) configured to store data.