A multi-exposure image registration method, device and equipment for adaptive block-by-block displacement estimation

Abstract

The present invention relates to a multi-exposure image registration method, device and equipment for adaptively estimating displacement by block, which belongs to the field of computer vision; the method includes obtaining at least two images with different exposure amounts, and determining a reference image and a source image therefrom; adaptively blocking the source image according to the gradient intensity variance, dividing it into a plurality of block images; locally matching the adaptive blocking of the source image and the reference image to estimate spatial transformation parameters; and mapping the source image to the space of the reference image according to the spatial transformation parameters to obtain the registered source image. The present invention can properly control the granularity of local displacement and improve computational efficiency. The present invention maps from maximum inter-class variance binarization to pixel value sorting binarization, and utilizes bidirectional displacement testing in two directions to further improve the robustness of the local displacement vector. The present invention has the advantages of low computational complexity, high speed, and easy implementation.

Description

Technical Field

[0001] The present invention relates to a multi-exposure image registration method, device and equipment for adaptive block displacement estimation, and belongs to the field of computer vision. Background Art

[0002] Existing imaging devices are often unable to directly capture high-dynamic-range brightness information in natural scenes. Driven by advances in computer vision, high-dynamic-range imaging (HDR) technology, through multiple exposures, can, to a certain extent, address these limitations. It is currently widely used in mobile phone cameras, security surveillance, smart cars, and other fields. A primary implementation of HDR technology is to fuse multiple low-dynamic-range images into a single HDR image containing sufficient HDR information, a process known as multi-exposure image fusion. During the capture of a multi-exposure image sequence, there is often relative motion between the camera and the scene. Direct image fusion inevitably produces artifacts, necessitating registration of the multi-exposure image sequence.

[0003] Image registration methods can be broadly categorized into two types: block-based methods and feature-point-based methods. The former utilizes spatial or transform-domain similarity between image blocks for estimation. Because they rely directly on the pixel values of the image blocks, they are significantly affected by lighting, noise, and deformation, and can easily lead to registration failure in complex applications. The latter extracts stable and robust feature points from images and then establishes spatial transformation relationships between images based on the feature point coordinates. This allows for relatively stable image registration in complex applications, but it also suffers from higher time complexity.

[0004] Image block-based registration methods, because they are essentially a measure of similarity, are prone to estimation errors when subject to interference such as illumination and noise. In addition, when there is a large deformation between image blocks, image block-based registration methods cannot achieve good results locally. For example, the MTB algorithm (Ward G. Fast, robust image registration for compositing high dynamic range photographs from hand-held exposures [J]. Journal of graphics tools, 2003, 8 (2): 17-30.) cannot solve the problem of multi-exposure image registration under local displacement. Feature point-based methods include feature point extraction, feature descriptor estimation, and feature point matching. The feature point detection and matching search process often requires a large amount of time, which is unbearable for terminal devices. Moreover, since terminal devices often move little during multi-exposure image capture in high dynamic range imaging applications, directly using feature point-based registration is not the best choice. Summary of the Invention

[0005] The present invention aims to provide a multi-exposure image registration method, device, and apparatus for mobile terminal applications using adaptive block-based displacement estimation. Considering the need for high dynamic range imaging in mobile terminals such as mobile phones, the displacement and deformation between multi-exposure images are relatively small, while computational resources and user experience require low time consumption. To achieve the best possible local alignment of multi-exposure images, the present invention proposes a multi-exposure image registration method, device, and apparatus using adaptive block-based displacement estimation.

[0006] The improvements of the present invention primarily include: Proposing a method for calculating local displacement between images using adaptive blocking. Adaptive blocking utilizes the variance of gradients within image blocks to discriminate, making blocks larger in areas with smaller gradient variances and smaller in areas with larger gradient variances. Local displacement between images is determined by comparing binary images between blocks, and blocks are screened based on the magnitude of image gradients within the blocks. A two-way verification method is also used to ensure the validity of the displacement.

[0007] In a first aspect of the present invention, the present invention provides a multi-exposure image registration method for adaptive block-by-block displacement estimation, the method comprising:

[0008] Acquire at least two images with different exposure amounts, and determine a reference image and a source image therefrom;

[0009] Adaptively dividing the source image into multiple block images according to the gradient intensity variance;

[0010] Performing local matching on the adaptive blocks of the source image and the reference image to estimate spatial transformation parameters;

[0011] The source image is mapped to the space of the reference image according to the spatial transformation parameters to obtain a registered source image.

[0012] In a second aspect of the present invention, the present invention further provides a multi-exposure image registration device for adaptively estimating displacement in blocks, the device comprising:

[0013] An acquisition module, configured to acquire at least two images with different exposure amounts and determine a reference image and a source image therefrom;

[0014] A partitioning module, configured to adaptively partition the source image into multiple partitioned images according to the gradient intensity variance;

[0015] an estimation module, configured to perform local matching between the adaptive blocks of the source image and the reference image to estimate spatial transformation parameters;

[0016] A registration module is used to map the source image to the space of the reference image according to the spatial transformation parameters to obtain a registered source image.

[0017] In a third aspect of the present invention, the present invention further provides a multi-exposure image registration method for adaptive block-by-block displacement estimation, comprising:

[0018] processor;

[0019] a memory for storing processor-executable instructions;

[0020] Wherein, the processor is configured to: execute the image registration method as described in the first aspect of the present invention.

[0021] Beneficial effects of the present invention:

[0022] This invention, aimed at mobile terminal applications, proposes a multi-exposure image registration method, apparatus, and device for adaptive block-based displacement estimation. The core concept is to adaptively partition the image into blocks based on the target content of the image to be registered. Local displacement vectors are then estimated for each block in parallel using a fast MTB algorithm. This method enables rapid multi-exposure image registration and can handle more complex spatial transformations than traditional global displacement methods. This adaptive block-based method appropriately controls the granularity of local displacements. When the image contains a lot of texture information, finer local block-based partitioning is possible, while larger local block-based partitioning is possible. Furthermore, when the pixel value variation within a block is small, the block is marked as invalid and excluded from the displacement vector calculation. For the calculation of local displacement vectors, the invention maps from maximum inter-class variance binarization to pixel value sorting binarization, utilizing bidirectional displacement checks in two directions to further enhance the robustness of the local displacement vectors. Compared to feature point-based registration methods, this method offers advantages such as reduced computational effort, high speed, and ease of implementation, making it widely applicable to mobile terminal registration scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 This is a flow chart of a multi-exposure image registration method for adaptive block-by-block displacement estimation according to an embodiment of the present invention;

[0024] Figure 2 is a flowchart of an adaptive block partitioning embodiment of the present invention;

[0025] Figure 3 1 is a schematic diagram of adaptive blocking according to an embodiment of the present invention;

[0026] Figure 4 is a local matching flow chart of an embodiment of the present invention;

[0027] Figure 5 The present invention is a schematic structural diagram of a multi-exposure image registration device with adaptive block-by-block displacement estimation according to an embodiment of the present invention. DETAILED DESCRIPTION

[0028] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0029] In order to make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention are further described in detail with reference to the accompanying drawings.

[0030] Image registration can be understood as the process of matching and overlaying two or more images acquired at different times, using different sensors (imaging devices, such as cameras), or under different conditions (such as weather, illumination, camera position, and angle). During image registration, the image to be registered can be mapped to the image space corresponding to the reference image by performing perspective transformation based on the mapping relationship between the reference image and the image to be registered. This allows the image to be registered with the reference image.

[0031] It is known from related technologies that in the process of image registration, the displacement and deformation between multi-exposure images are often relatively small, and the computing resources and user experience require low time consumption, but it is not possible to register multi-exposure images well and quickly. In order to solve the problem that the MTB algorithm can only handle multi-exposure image registration under global displacement, the present invention provides a multi-exposure image registration method with adaptive block estimation displacement, and proposes a method for calculating the local displacement between images using adaptive block, so as to obtain spatial transformation parameters. Figure 1 As shown, the present invention mainly includes four steps, namely inputting the image to be registered, adaptively dividing the image into blocks, establishing the matching point set and outputting the registered image, which are described in detail below.

[0032] 101. Acquire at least two images with different exposures, and determine a reference image and a source image therefrom;

[0033] In one embodiment, at least two images with different exposures are acquired. The at least two images with different exposures may be multiple frames capturing the same scene or the same target object. Furthermore, a reference image may be determined from the multiple frames. In one example, the reference image may be the image with the highest resolution. In another example, the image to be registered may be any image in the multiple frames other than the reference image.

[0034] In the technical solution description of the present invention, two images with different exposure values are used as an example for registration, and the brightness channel of the image is extracted in the color space, and the brightness channel of the image with the larger normal exposure area is recorded as the reference image I.ref , the image brightness channel with smaller normal exposure area is the source image I ref , taking the source image as the image to be registered, and utilizing the mapping relationship between the reference image and the image to be registered, the image to be registered is processed with perspective change to map the image to be registered to the image space corresponding to the reference image, thereby realizing the registration of the image to be registered with the reference image.

[0035] 102. Adaptively divide the source image into multiple block images according to the gradient intensity variance;

[0036] In an embodiment of the present invention, considering that although the scenes between the images to be registered are basically the same, the difference in exposure time may cause inconsistency in the effective information between different images, in order to reduce unnecessary calculations and improve time efficiency, the present invention chooses to use an image with fewer normal exposure areas, namely the source image, to implement an adaptive blocking operation.

[0037] The technical process of adaptive segmentation is as follows Figure 2 Specifically, it may include:

[0038] 211. Evenly divide the source image into multiple block images of a certain scale;

[0039] In an embodiment of the present invention, the source image may first be evenly divided into blocks of a certain size. For example, the source image may be divided into four block images of a size of 2×2, where each block image has the same size. For example, if the source image is 256×256 in size, after being divided into 2×2 blocks, the size of each block image is 128×128. These four block images represent four sub-images of the source image in the upper, lower, left, and right directions, respectively.

[0040] 212. Calculate the gradient strength and gradient strength variance of each of the block images;

[0041] In an embodiment of the present invention, the calculation of gradient strength can use horizontal, vertical or diagonal convolution kernels, and the gradient variance within the image block can use a fast algorithm. Moreover, since each image block is independent and does not affect each other, it can be performed through parallel computing to further improve time efficiency.

[0042] 213. Determine whether the gradient strength variance of each of the block images is greater than a first threshold.

[0043] In an embodiment of the present invention, the richness of the texture information of each block image can be measured by calculating the variance of the gradient intensity of each block image. Therefore, this embodiment determines the texture of the block image by judging the difference between the variance of the gradient intensity of the block image and the first threshold; wherein, the first threshold can be set according to the experience of those skilled in the art, and the present invention does not make any specific limitation on this.

[0044] 214. If the value is less than the first threshold, retain the divided image;

[0045] 215. If it is not less than the first threshold, continue dividing the block image into block images of smaller sizes until the minimum block size is reached, and end the adaptive iterative process.

[0046] It is understandable that the present invention believes that when image texture information is rich, smaller image blocks can better reflect local information, while when texture information is scarce, larger image blocks can better ensure that the information within each image block is valid. The image gradient intensity and variance can, to a certain extent, represent the richness of texture information. Therefore, the present invention iteratively obtains the block results based on gradient information. The present invention determines whether to continue block segmentation by determining the relationship between the gradient intensity variance of the block image and a first threshold value, thereby dividing images containing more texture information into finer-grained block images and images containing less texture information into relatively coarse-grained block images, thereby displaying the texture information of the source image using block images of different granularities.

[0047] In a preferred embodiment of the present invention, the above steps are further described in detail with reference to specific data:

[0048] First, the source image I src Divide it into 2×2 blocks evenly, which can be recorded as {Pa i}, where i = 1, 2, 3, 4. The block operation is only to divide the coordinates in the coordinate space of the image, and does not actually segment the image. Then, the gradient strength of the image in each block is calculated and recorded as And calculate the variance of the gradient strength within each block Where i=1,2,3,4.

[0049] For the blocks {Pa i}, judge the gradient variance of the image in each block one by one Is it greater than a certain threshold σ t If a block Pa i If the condition is met, the block is divided evenly into blocks, which can be recorded as {Pa ij}, where j = 1, 2, 3, 4, and continue to calculate the gradient strength of the image in each block and gradient variance Then judge the gradient variance one by one Is it greater than a certain threshold σ t , and repeat this operation. iIf the condition is not met, the current block is retained. In actual operation, in order to prevent the loop from falling into invalid block subdivision, a minimum block size is set to end the iteration process early.

[0050] like Figure 3 As shown in the figure, taking the source image size of 256×256 as an example, after the source image is adaptively divided into blocks, each block image may be further subdivided or may not need to be subdivided due to the different texture information contained in it. Therefore, the source image can be presented as block images of various sizes. For example, the block image in the lower left corner has been divided once, and its size is 128×128, while other block images may be divided multiple times. For example, the block image in the upper left corner has been divided twice, and its size is 64×64, and the block image in the lower right corner has been divided three times, and its size is 32×32. This division method can not only ensure that the information in each block image is valid, but also reflect local information as much as possible, and has relatively small computing resource consumption and time consumption, which is beneficial to image registration scenarios of mobile terminals.

[0051] In a preferred embodiment of the present invention, the adaptive blocking technical process further includes step 216:

[0052] 216. Screen the image blocks; calculate the average value and standard deviation of the gradient strength in all image blocks; extract the image blocks whose gradient strength is less than the difference between the average value and the standard deviation.

[0053] After the above iterative operation, the block results of different sizes can be obtained adaptively according to the image content. In order to further improve the robustness, the present invention continues to implement the block screening operation to further eliminate the blocks with less gradient changes. The blocks obtained after the adaptive block division are denoted as {Pa k}, where k = 1, 2, ..., K, there are K blocks in total, and the gradient strength of the image in each block is recorded as Calculate the average value μ of the image gradient strength in all blocks Gsrc and standard deviation σ Gsrc , and then directly remove the image whose gradient intensity value is less than μ Gsrc -σ Gsrc The final effective adaptive block {Pa m}, where m = 1, 2, …, M, with a total of M blocks.

[0054] 103. Perform local matching on the adaptive blocks of the source image and the reference image to estimate spatial transformation parameters. In this embodiment of the present invention, Figure 4 As shown, the local matching includes: using the adaptive block {Pa m} as a unit, respectively in the reference image I refand source image I src The maximum inter-class variance binarization, gray value sorting binarization, and bidirectional displacement test are performed on the image, and the spatial transformation parameters between the reference image and the source image are estimated based on the obtained matching point set. The specific description is as follows.

[0055] 311. Perform maximum inter-class variance binarization on the reference image and the source image respectively, and calculate the binarized images of the reference image and the source image in all blocks and the binarized segmentation threshold;

[0056] In the embodiment of the present invention, the maximum inter-class variance binarization, namely the Otsu algorithm (Otsu NA threshold selection method from gray-level histograms [J]. IEEE transactions on systems, man, and cybernetics, 1979, 9 (1): 62-66.) is a common image fast binarization method, the core of which is to find a binary threshold that maximizes the inter-class variance. The present invention directly uses the Otsu algorithm to calculate the reference image I ref and source image I re f in all blocks {Pa m Binarized image in and That is, the reference image I ref and source image I src The maximum inter-class variance binarized image and the binarized segmentation threshold and Where m = 1, 2,…, M.

[0057] 312. Binarize the pixel values of the reference image and the source image respectively, and determine the pixel value-ordered binarized images of all blocks using the corresponding binarization segmentation thresholds;

[0058] In the embodiment of the present invention, for the reference image I ref , count the binary segmentation thresholds in all blocks The first position ratio of the positive sequence of image pixel values in the corresponding block If a binary segmentation threshold If there are multiple identical pixel values in the image, the average value is taken when calculating the position ratio in the positive order. After that, in the source image I src In all blocks, find the value in the positive sequence of the image pixel value corresponding to the ratio And use this as the binary segmentation threshold to get the source image I src Pixel value sorting binary image

[0059] In the embodiment of the present invention, for the source image I src , count the binary segmentation thresholds in all blocks The second position ratio in the positive order of the image pixel values in the corresponding block If a binary segmentation threshold If there are multiple identical pixel values in the image, the average value is taken when calculating the position ratio in the positive order. After that, in the reference image I ref In all blocks, find the value in the positive sequence of the image pixel value corresponding to the ratio And use this as the binary segmentation threshold to get the reference image I ref Pixel value sorting binary image

[0060] 313. Using the maximum inter-class variance binary image and the pixel value sorting binary image to perform bidirectional displacement testing, multiple valid displacement vectors are determined;

[0061] In the embodiment of the present invention, the maximum inter-class variance is used to binarize the image. Binarized image by sorting pixel values Implement the MTB algorithm to obtain M pairs of displacement vectors Then use the pixel value to sort the binary image and the maximum inter-class variance binarized image Implement the MTB algorithm to obtain M pairs of displacement vectors Where m=1,2,…,M. Respectively, from the reference image I ref The maximum inter-class variance binarization threshold is mapped to the source image I src Between the binary images formed on the M blocks {Pa m}, and the M displacement vectors from the source image I src The maximum inter-class variance binarization threshold is mapped to the reference image I ref Between the binary images formed on the M blocks {Pa m}M displacement vectors on .

[0062] In a preferred embodiment of the present invention, in order to improve the robustness of the displacement vector, the present invention performs a bidirectional check on the displacement vectors in the two cases. The specific operation method is: traverse and calculate the mth displacement vector and When the angle between t When , calculate the average of the two current displacement vectors and keep it, otherwise remove the current displacement vector. Finally, get Z valid displacement vectors And the corresponding Z blocks {Pa z}, where z = 1, 2,…, Z.

[0063] 314. According to the effective displacement vector, the spatial transformation parameters are estimated.

[0064] In the embodiment of the present invention, in the image space, Z blocks of different sizes {Pa z}, and retains the reference image I in each block ref and source image I src Displacement vector between It characterizes the local alignment of the images to be registered, and based on this, it can estimate the spatial transformation parameters. Typically, classic techniques such as multi-homography matrix method, optical flow method, grid optimization method, etc. can be used, or combined with deep learning methods.

[0065] It is understood that the order of steps 311 and 312 can be swapped. For ease of description, step 311 is placed first and step 312 is placed second. In practice, the order of the two steps can be arbitrary. 104. Map the source image to the space of the reference image based on the spatial transformation parameters to obtain a registered source image.

[0066] In the embodiment of the present invention, the source image I is converted into src Mapped to the reference image I ref space, and obtain the source image I′ after registration src .

[0067] Figure 5 FIG. 1 is a schematic structural diagram of a multi-exposure image registration device for adaptive block-wise displacement estimation according to an embodiment of the present invention. Figure 5 As shown, the device includes:

[0068] An acquisition module, configured to acquire at least two images with different exposure amounts and determine a reference image and a source image therefrom;

[0069] A partitioning module, configured to adaptively partition the source image into multiple partitioned images according to the gradient intensity variance;

[0070] an estimation module, configured to perform local matching between the adaptive blocks of the source image and the reference image to estimate spatial transformation parameters;

[0071] A registration module is used to map the source image to the space of the reference image according to the spatial transformation parameters to obtain a registered source image.

[0072] In yet another embodiment of the present invention, the present invention further proposes a multi-exposure image registration device for adaptive block-by-block displacement estimation, comprising:

[0073] processor;

[0074] a memory for storing processor-executable instructions;

[0075] Wherein, the processor is configured to: execute the image registration method as described in the present invention.

[0076] It is understandable that the image registration device provided by the embodiment of the present disclosure includes hardware structures and / or software modules corresponding to the execution of each function in order to realize the above functions. In combination with the units and algorithm steps of each example disclosed in the embodiment of the present disclosure, the embodiment of the present disclosure can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the technical solution of the embodiment of the present disclosure.

[0077] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the relevant hardware through a program, and the program can be stored in a computer-readable storage medium, which may include: ROM, RAM, disk or CD, etc.

[0078] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A multi-exposure image registration method with adaptive block-by-block displacement estimation, characterized in that: The method comprises: Acquire at least two images with different exposure amounts, and determine a reference image and a source image therefrom; adaptively divide the source image into multiple block images according to the gradient intensity variance; The adaptive blocking includes: Evenly dividing the source image into multiple block images of a certain scale; Calculating the gradient strength and gradient strength variance of each of the block images; Determine whether the gradient strength variance of each of the block images is greater than a first threshold, If it is less than the first threshold, retaining the block image; If it is not less than the first threshold, continue dividing the block image into smaller block images until the minimum block size is reached, and end the adaptive iterative process; Performing local matching on the adaptive blocks of the source image and the reference image to estimate spatial transformation parameters; The local matching includes: Perform maximum inter-class variance binarization on the reference image and the source image respectively, and calculate the binarized images of the reference image and the source image in all blocks as well as the binarized segmentation threshold; The reference image and the source image are binarized by sorting the pixel values, and the corresponding binarization segmentation threshold is used to determine the pixel value sorting binarized images of all blocks; The maximum inter-class variance binary image and the pixel value sorting binary image are used to perform bidirectional displacement test and determine multiple valid displacement vectors. According to the effective displacement vector, the spatial transformation parameters are estimated; The bidirectional displacement test includes: A plurality of pairs of first displacement vectors are obtained by using a maximum inter-class variance binarized image of a reference image and a pixel value sorted binarized image of a source image; A plurality of pairs of second displacement vectors are obtained by sorting a binary image of a reference image and a maximum inter-class variance binary image of a source image using pixel values of the reference image; traversing and calculating the angles between multiple pairs of first displacement vectors and multiple pairs of second displacement traversal vectors, and when the angle is less than a second threshold, calculating the average of the current two displacement vectors and retaining the average; otherwise, discarding the current displacement vector; until multiple valid displacement vectors are determined; The source image is mapped to the space of the reference image according to the spatial transformation parameters to obtain a registered source image.

2. The multi-exposure image registration method with adaptive block-by-block displacement estimation according to claim 1, characterized in that: The adaptive blocking also includes screening the blocked images; calculating the average value and standard deviation of the gradient strength in all image blocks; and extracting image blocks whose gradient strength is less than the difference between the average value and the standard deviation.

3. The multi-exposure image registration method with adaptive block-by-block displacement estimation according to claim 1, characterized in that: The pixel value sorting binarization includes: For the reference image, the ratio of the first position of the binary segmentation threshold in the positive sequence of the image pixel values in the corresponding image block is counted; In all blocks of the source image, searching for a value in the positive sequence of image pixel values corresponding to the first position ratio, and using this value as a binarization segmentation threshold to obtain a pixel value-ordered binary image of the source image; For the source image, the proportion of the second position of the binary segmentation threshold in the positive sequence of the image pixel values in the corresponding image blocks is counted; In all blocks of the reference image, the value in the positive sequence of the image pixel values corresponding to the second position ratio is found, and this is used as a binarization segmentation threshold to obtain a pixel value sorted binary image of the reference image.

4. The multi-exposure image registration method with adaptive block-by-block displacement estimation according to claim 3, characterized in that: If a certain binary segmentation threshold corresponds to multiple identical pixel values, the average value is taken when calculating the position ratio in the positive sequence.

5. A multi-exposure image registration device with adaptive block displacement estimation, applied to the multi-exposure image registration method with adaptive block displacement estimation according to any one of claims 1 to 4, characterized in that: The device comprises: An acquisition module, configured to acquire at least two images with different exposure amounts and determine a reference image and a source image therefrom; A partitioning module, configured to adaptively partition the source image into multiple partitioned images according to the gradient intensity variance; an estimation module, configured to perform local matching between the adaptive blocks of the source image and the reference image to estimate spatial transformation parameters; A registration module is used to map the source image to the space of the reference image according to the spatial transformation parameters to obtain a registered source image.

6. A multi-exposure image registration device with adaptive block-by-block displacement estimation, characterized in that: include: processor; a memory for storing processor-executable instructions; Wherein, the processor is configured to: execute the image registration method according to any one of claims 1 to 4.

Images