Image stitching method and device, electronic device, and storage medium

Through the combination of feature point algorithm and global feature inheritance algorithm, the image instability and stitching error problems in stitching images in in vitro three-dimensional biological model imaging are solved, which improves the stability and accuracy of stitching and enhances the stitching quality.

CN118570061BActive Publication Date: 2024-11-26JIANGSU AVATARGET BIOTECHNOLOGY CO LTD
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN202411053517.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-02
Publication Date
2024-11-26
Estimated Expiration
2044-08-02

AI Technical Summary

Technical Problem

The prior art has image instability and stitching errors in in vitro three-dimensional biological model imaging images, especially in the case of high connotation instrument jitter and accuracy errors, and the existing algorithms are efficient when there is no obvious feature or large-scale image processing. Low, poor accuracy.

Method used

The feature point algorithm is used for feature point detection and matching, the splicing amount is determined based on the matching results, and the global feature inheritance algorithm is used to make up for it when the feature point matching fails, and the image splicing is combined with the preset overlap ratio and the weighted fusion algorithm.

Benefits of technology

Improves the stability and accuracy of image stitching, reduces the time-consuming feature matching calculation, and enhances the stitching quality, especially in the absence of obvious targets or mostly blank areas.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118570061B_ABST
    Figure CN118570061B_ABST
Patent Text Reader

Abstract

The present disclosure relates to the field of image stitching, and provides an image stitching method and device, an electronic device, and a storage medium. The method includes: obtaining a group of images to be stitched that are shot according to a preset overlap ratio; using a feature point algorithm to detect feature points of each image to be stitched, and obtaining corresponding feature point information; using a feature point matching algorithm to match feature points of two adjacent images to be stitched based on the feature point information, and obtaining a matching result; if the matching result indicates a successful match, determining the stitching amount corresponding to the two adjacent images to be stitched according to the distance value of the successfully matched feature point pair; if the matching result indicates a failed match, determining the stitching amount corresponding to the two adjacent images to be stitched using a global feature inheritance algorithm; stitching and fusing each image to be stitched according to the stitching amount, and obtaining a target stitching image. The present disclosure has a faster running speed, a good effect in dealing with random errors, and a better effect when there are no obvious features in the stitching area.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present disclosure relates to the technical field of image stitching, and in particular to a method and device, electronic equipment, and storage medium for stitching images of an in vitro three-dimensional biological model. Background Art

[0002] In vitro three-dimensional organ models have been booming in recent years. For example, models of lungs, kidneys, intestines, livers, hearts, blood vessels, skins, brains, etc. are constructed on chips. By combining methods from multiple disciplines such as cell biology, engineering, and biomaterials, various living cells, tissues, and organ microenvironments are simulated in vitro to reflect the main structural and functional characteristics of human tissues and organs. For example, organoids constructed from tissue extraction or organoids differentiated from stem cells are a type of in vitro three-dimensional organ model.

[0003] Regularly observing in vitro 3D organ models such as organoids in microscopic images to obtain morphological or growth characteristics is a necessary condition for organ drug screening. Usually, when using a high-power microscope to shoot in vitro 3D organ models, image stitching is required due to the need to expand the field of view.

[0004] At present, the main frame of the high-content instrument is usually supported by the base. When the environment where the base is located shakes, the imaging module in the high-content instrument will also shake, resulting in an unstable state of the image captured by the imaging module. When the image is stitched on this basis, the puzzle error is large. In addition, the high-content instrument itself usually has a certain accuracy error. For example, each time the stage is moved, the motor starts and causes a slight vibration in the high-content instrument, resulting in an unstable image captured by the high-content instrument, which in turn causes image stitching errors. When the above two situations are superimposed on each other, the error in the image stitching effect of the high-content instrument will become larger, resulting in poor quality of the stitched image. At the same time, there may be no obvious target in the image captured under the microscope field of view, or most of the area may be blank, which also adds difficulty to achieving accurate puzzles.

[0005] With the development of image stitching technology, image stitching algorithms based on feature matching and image stitching algorithms based on region matching have emerged in the existing technology. However, image stitching algorithms based on feature matching, such as the Scale-invariant feature transform (SIFT) algorithm and the Speeded Up Robust Features (SURF) algorithm, not only have high time costs and slow feature matching speeds, but also cannot cope with situations where there are no obvious features in the image. Image stitching algorithms based on region matching not only have long calculation times and a single matching stitching mode, but also cannot cope with situations where there are no obvious signals in the image, are sensitive to scale changes, and are not suitable for large-scale images. Summary of the invention

[0006] The present disclosure aims to solve at least one of the problems existing in the prior art and provides an image stitching method and device, an electronic device, and a storage medium.

[0007] In one aspect of the present disclosure, there is provided an image stitching method, the image stitching method comprising:

[0008] Acquire a set of images to be stitched that are taken according to a preset overlap ratio;

[0009] Using a feature point algorithm to detect feature points of each of the images to be spliced, to obtain feature point information corresponding to each of the images to be spliced;

[0010] Based on the feature point information, feature points of two adjacent images to be stitched are matched using a feature point matching algorithm to obtain corresponding matching results;

[0011] If the matching result indicates that the feature points of the two adjacent images to be stitched are matched successfully, then determining the stitching amount corresponding to the two adjacent images to be stitched according to the distance value of the successfully matched feature point pair;

[0012] If the matching result indicates that the feature points of the two adjacent images to be stitched fail to match, a global feature inheritance algorithm is used to determine the stitching amount corresponding to the two adjacent images to be stitched;

[0013] According to the stitching amount, each of the images to be stitched is stitched and fused to obtain a target stitching image.

[0014] Optionally, the feature point information includes feature point coordinates and a feature description matrix; the feature point algorithm includes a directed ring feature point algorithm;

[0015] The step of using a feature point algorithm to detect feature points of each of the images to be stitched to obtain feature point information corresponding to each of the images to be stitched includes:

[0016] A circular sliding window with a preset diameter is used to perform feature point detection on each of the images to be stitched, and the center coordinates of the circular sliding window are used as the feature point coordinates. The change gradients of all pixel points in the circular sliding window and their adjacent pixel points are calculated to generate a corresponding gradient size matrix. The contour with the maximum gradient change in the gradient size matrix is ​​marked to obtain the feature description matrix.

[0017] Optionally, matching the feature points of two adjacent images to be stitched using a feature point matching algorithm based on the feature point information to obtain corresponding matching results includes:

[0018] For two adjacent images to be stitched, the feature description matrices corresponding to one of the images to be stitched are sequentially used as target feature point information matrices, and the feature description matrices corresponding to the other image to be stitched are respectively used as feature point information matrices to be matched;

[0019] Subtracting each of the target feature point information matrices from each of the feature point information matrices to be matched one by one in turn, and calculating the projection difference value between the maximum gradient change contour in each of the target feature point information matrices and the maximum gradient change contour in each of the feature point information matrices to be matched;

[0020] Determine whether the minimum projection difference value corresponding to the target feature point information matrix is ​​greater than a preset threshold: if so, the matching result indicates that the feature point matching fails; if not, the matching result indicates that the feature point matching succeeds.

[0021] Optionally, determining the stitching amount corresponding to the two adjacent images to be stitched according to the distance values ​​of the successfully matched feature point pairs includes:

[0022] According to the absolute pixel coordinates of the midline points of each feature point on the images to be stitched, respectively calculating the distance between the two feature points included in each feature point pair, to obtain a plurality of feature point pair distances corresponding to two adjacent images to be stitched;

[0023] The value with the largest number of occurrences among the distances between the plurality of feature point pairs is used as the stitching amount corresponding to the two adjacent images to be stitched.

[0024] Optionally, the using a global feature inheritance algorithm to determine the stitching amount corresponding to the two adjacent images to be stitched includes:

[0025] Based on the stitching position of the two adjacent images to be stitched where the feature point matching fails, searching around for other two adjacent images to be stitched that have the same stitching direction as the two adjacent images to be stitched and have successful feature point matching;

[0026] The stitching amount corresponding to the two adjacent images to be stitched whose feature points fail to match is determined according to the feature point information corresponding to the other two adjacent images to be stitched that are searched.

[0027] Optionally, the step of performing stitching and fusing the images to be stitched according to the stitching amount to obtain a target stitched image includes:

[0028] Determining, according to the stitching amounts corresponding to the two adjacent images to be stitched, image overlapping areas corresponding to the two adjacent images to be stitched;

[0029] The weights are determined by a logarithmic function, and the pixels in the image overlap region corresponding to two adjacent images to be stitched are superimposed and fused according to the weights to obtain the target stitched image.

[0030] Optionally, the number b of the images to be stitched satisfies: b≥(ad+x) / (dx); wherein a represents the number of images to be stitched, d represents the field of view size of a single image to be stitched, and x represents the preset overlap ratio.

[0031] Another aspect of the present disclosure provides a method for stitching images of an in vitro three-dimensional biological model, which is applied to in vitro three-dimensional biological model imaging images and includes the image stitching method described above.

[0032] Another aspect of the present disclosure provides an image stitching device, the image stitching device comprising:

[0033] An acquisition module, used for acquiring a set of images to be stitched that are shot according to a preset overlap ratio;

[0034] A detection module, used to perform feature point detection on each of the images to be spliced ​​using a feature point algorithm to obtain feature point information corresponding to each of the images to be spliced;

[0035] A matching module, used to match the feature points of two adjacent images to be stitched using a feature point matching algorithm based on the feature point information to obtain a corresponding matching result;

[0036] A determination module, configured to: if the matching result indicates that the feature points of the two adjacent images to be stitched are matched successfully, determine the stitching amount corresponding to the two adjacent images to be stitched according to the distance value of the successfully matched feature point pair; if the matching result indicates that the feature points of the two adjacent images to be stitched fail to match, determine the stitching amount corresponding to the two adjacent images to be stitched using a global feature inheritance algorithm;

[0037] The stitching and fusion module is used to stitch and fuse the images to be stitched according to the stitching amount to obtain a target stitching image.

[0038] Optionally, the feature point information includes feature point coordinates and a feature description matrix; the feature point algorithm includes a directed ring feature point algorithm;

[0039] The detection module is used to detect feature points of each of the images to be spliced ​​using a feature point algorithm to obtain feature point information corresponding to each of the images to be spliced, including:

[0040] The detection module is used to: use a circular sliding window with a preset diameter to perform feature point detection on each of the images to be stitched, use the center coordinates of the circular sliding window as the feature point coordinates, calculate the change gradients of all pixel points in the circular sliding window and their adjacent pixel points, generate a corresponding gradient size matrix, mark the contour with the maximum gradient change in the gradient size matrix, and obtain the feature description matrix.

[0041] Optionally, the matching module is used to match the feature points of two adjacent images to be stitched using a feature point matching algorithm based on the feature point information to obtain a corresponding matching result, including:

[0042] The matching module is used for:

[0043] For two adjacent images to be stitched, the feature description matrices corresponding to one of the images to be stitched are sequentially used as target feature point information matrices, and the feature description matrices corresponding to the other image to be stitched are respectively used as feature point information matrices to be matched;

[0044] Subtracting each of the target feature point information matrices from each of the feature point information matrices to be matched one by one in turn, and calculating the projection difference value between the maximum gradient change contour in each of the target feature point information matrices and the maximum gradient change contour in each of the feature point information matrices to be matched;

[0045] Determine whether the minimum projection difference value corresponding to the target feature point information matrix is ​​greater than a preset threshold: if so, the matching result indicates that the feature point matching fails; if not, the matching result indicates that the feature point matching succeeds.

[0046] Optionally, the determination module is used to determine the stitching amount corresponding to the two adjacent images to be stitched according to the distance values ​​of the successfully matched feature point pairs, including:

[0047] The determining module is used to:

[0048] According to the absolute pixel coordinates of the midline points of each feature point on the images to be stitched, respectively calculating the distance between the two feature points included in each feature point pair, to obtain a plurality of feature point pair distances corresponding to two adjacent images to be stitched;

[0049] The value with the largest number of occurrences among the distances between the plurality of feature point pairs is used as the stitching amount corresponding to the two adjacent images to be stitched.

[0050] Optionally, the determining module is used to determine the stitching amount corresponding to the two adjacent images to be stitched by using a global feature inheritance algorithm, including:

[0051] The determining module is used to:

[0052] Based on the stitching position of the two adjacent images to be stitched where the feature point matching fails, searching around for other two adjacent images to be stitched that have the same stitching direction as the two adjacent images to be stitched and have successful feature point matching;

[0053] The stitching amount corresponding to the two adjacent images to be stitched whose feature points fail to match is determined according to the feature point information corresponding to the other two adjacent images to be stitched that are searched.

[0054] Optionally, the stitching and fusion module is used to stitch and fuse the images to be stitched according to the stitching amount to obtain a target stitching image, including:

[0055] The splicing and fusion module is used for:

[0056] Determining, according to the stitching amounts corresponding to the two adjacent images to be stitched, image overlapping areas corresponding to the two adjacent images to be stitched;

[0057] The weights are determined by a logarithmic function, and the pixels in the image overlap region corresponding to two adjacent images to be stitched are superimposed and fused according to the weights to obtain the target stitched image.

[0058] Optionally, the number b of the images to be stitched satisfies: b≥(ad+x) / (dx); wherein a represents the number of images to be stitched, d represents the field of view size of a single image to be stitched, and x represents the preset overlap ratio.

[0059] Another aspect of the present disclosure provides an electronic device, including:

[0060] at least one processor; and,

[0061] a memory communicatively connected to at least one processor; wherein,

[0062] The memory stores instructions that can be executed by at least one processor, and the instructions are executed by at least one processor so that the at least one processor can execute the image stitching method described above or the image stitching method of in vitro three-dimensional biological model imaging.

[0063] Another aspect of the present disclosure provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the image stitching method described above or the image stitching method of an in vitro three-dimensional biological model.

[0064] Compared with the prior art, the present invention uses images to be spliced ​​that are shot according to a preset overlap ratio for splicing, which can ensure that the approximate range of the splicing amount is stable and improve the stability of image splicing. In view of the difficulty of uncertain errors in the image splicing process, a feature point algorithm is designed, and a feature point matching algorithm is used to determine the feature point pairs of two adjacent images to be spliced, and then the splicing amount of the two adjacent images to be spliced ​​is determined according to the distance value of the feature point pairs. This can not only quickly detect the feature point information of the images to be spliced, but also retain the directionality of the feature point information, improve the accuracy of feature point matching, and reduce the time-consuming feature matching calculation. In view of the situation where there may be no obvious target or most of the blanks in the images to be spliced, resulting in feature detection and matching failure, a global feature inheritance algorithm is designed to achieve splicing compensation for feature detection and matching failure. By splicing and fusing each image to be spliced ​​according to the splicing amount, the image splicing quality is also improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] One or more embodiments are exemplarily described by pictures in the corresponding drawings, and these exemplifications do not constitute limitations on the embodiments. Elements with the same reference numerals in the drawings represent similar elements, and unless otherwise stated, the figures in the drawings do not constitute proportional limitations.

[0066] Figure 1 A flowchart of an image stitching method provided in one embodiment of the present disclosure;

[0067] Figure 2 A schematic diagram illustrating overlapping photography provided by another embodiment of the present disclosure;

[0068] Figure 3 A schematic diagram of the principle of a directed ring feature point algorithm provided by another embodiment of the present disclosure;

[0069] Figure 4 A schematic diagram of a feature point matching process provided by another embodiment of the present disclosure;

[0070] Figure 5 A flowchart of a microscopic image stitching method provided in another embodiment of the present disclosure;

[0071] Figure 6 A flowchart of an image stitching method provided in another embodiment of the present disclosure;

[0072] Figure 7 A schematic structural diagram of an image stitching device provided in another embodiment of the present disclosure;

[0073] Figure 8 A schematic structural diagram of an electronic device provided in another embodiment of the present disclosure. DETAILED DESCRIPTION

[0074] In order to make the purpose, technical scheme and advantages of the embodiments of the present disclosure clearer, the embodiments of the present disclosure will be described in detail below in conjunction with the accompanying drawings. However, it can be understood by those skilled in the art that in each embodiment of the present disclosure, many technical details are proposed in order to enable readers to better understand the present disclosure. However, even without these technical details and various changes and modifications based on the following embodiments, the technical scheme claimed for protection in the present disclosure can also be implemented. The division of the following embodiments is for the convenience of description and should not constitute any limitation on the specific implementation of the present disclosure. The various embodiments can be combined and referenced with each other without contradiction.

[0075] One embodiment of the present disclosure relates to an image stitching method, the process of which is as follows: Figure 1 As shown, including:

[0076] Step S110, obtaining a group of images to be stitched that are taken according to a preset overlap ratio.

[0077] Specifically, step S110 can use any instrument capable of taking pictures, such as a high-content instrument, to take a group of images to be stitched according to a preset overlap ratio. The preset overlap ratio refers to the ratio of the overlapping area of ​​two adjacent images to be stitched. In other words, when taking the images to be stitched, the latter image to be stitched needs to have a certain ratio of overlapping area with the previous image to be stitched, and the certain ratio is the preset overlap ratio.

[0078] Since the image stitching method provided in this embodiment needs to extract feature points from the images to be stitched in the subsequent steps, there needs to be a certain proportion of overlapping areas between adjacent images to be stitched to increase the success rate of feature point recognition. At the same time, considering that the instrument used for shooting may be jittery and feature points cannot be detected in the image, each pair of adjacent images to be stitched uses the same preset overlap ratio to ensure that the approximate range of the stitching amount is stable, which facilitates the subsequent steps to use the global feature inheritance algorithm to achieve global feature inheritance and improve the stability and accuracy of image stitching.

[0079] In addition, due to the existence of the preset overlap ratio, compared with the case where there is no overlap between two adjacent images to be stitched, more images to be stitched need to be shot to make up for the missing shooting area caused by the overlap. Figure 2 , assuming that there is no overlapping area between two adjacent images to be stitched, a set of images to be stitched that need to be taken constitutes a 3×4 image matrix, such as Figure 2 The image matrix shown by the expected stitching result marked by the dotted line in the middle, after setting the preset overlap ratio, the set of images to be stitched will form a 4×6 image matrix, such as Figure 2 The image matrix composed of the actual shooting trajectory marked by the solid line ( Figure 2 Only the 1st to 7th images to be stitched in the image matrix are shown).

[0080] Assuming that the target redundancy, that is, the preset overlap ratio, is x, the single field of view size, that is, the field of view size of a single image to be stitched, is d, and the number of images to be stitched is expected to be a, then when the images to be stitched are taken according to the preset overlap ratio, the number b of images to be stitched required must satisfy: bd-(b-1)×x-2x≥ad, that is, b≥(ad+x) / (dx).

[0081] According to actual tests, setting the value of the preset overlap ratio x to 0.25 can reduce the number of redundant photos, that is, the number of images to be stitched, while ensuring the quality of image stitching, thereby improving the execution speed of the image stitching method.

[0082] It should be noted that step S110 can directly shoot a group of images to be stitched according to a preset overlap ratio, or obtain a group of images to be stitched that have been shot in advance according to a preset overlap ratio. This embodiment does not limit this, as long as the acquired group of images to be stitched is shot according to the preset overlap ratio.

[0083] Step S120 , using a feature point algorithm to perform feature point detection on each image to be spliced, to obtain feature point information corresponding to each image to be spliced.

[0084] Specifically, the feature point algorithm may include, but is not limited to, Features from Accelerated Segment Test (FAST) detection algorithm, Scale-Invariant Feature Transform (SIFT) algorithm, Speeded Up Robust Features (SURF) algorithm, Harris corner detection algorithm, ORB (Oriented Fast and Rotated Brief) algorithm, Hessian-Laplace algorithm, directed ring feature point algorithm, etc. Step S120 may use any feature point algorithm to perform feature point detection on each image to be spliced ​​in turn to obtain corresponding feature point information. Generally, the image to be spliced ​​usually contains multiple feature points, so the feature point information corresponding to the image to be spliced ​​includes relevant information of each detected feature point.

[0085] Exemplarily, the feature point information corresponding to the images to be spliced ​​includes feature point coordinates and a feature description matrix. The feature point algorithm is preferably a directed ring feature point algorithm. Step S120 uses the feature point algorithm to detect feature points on each image to be spliced, including: using a circular sliding window of a preset diameter, respectively detecting feature points on each image to be spliced, using the center coordinates of the circular sliding window as the feature point coordinates, calculating the change gradients of all pixels in the circular sliding window and their adjacent pixels, generating a corresponding gradient size matrix, marking the contour with the maximum gradient change in the gradient size matrix, and obtaining a feature description matrix.

[0086] Specifically, the actual size of the preset diameter can be determined according to the specific type of the image to be stitched adopted by the image stitching method. For example, when the image to be stitched adopted by the image stitching method is an in vitro three-dimensional biological model imaging image, the range of the preset diameter is preferably 7 pixels to 13 pixels.

[0087] Combined Figure 3 , assuming that the preset diameter of the circular sliding window is 9 pixels, the coordinates of the center of the circular sliding window, that is, the coordinates of the pixel at the center of the circle, can be used as the center position coordinates of the detected feature points, that is, the feature point coordinates. Calculate the change gradients of each pixel in the circular sliding window and its adjacent pixels respectively to obtain the corresponding gradient change values, and then the gradient change values ​​corresponding to each pixel can form a gradient size matrix. In the gradient size matrix, select the contour with the largest gradient change, that is, the contour with the largest gradient change value, and mark it. Then the marked gradient size matrix can be used as a feature description matrix to describe the feature information in the area where the circular sliding window is currently located in the image to be spliced.

[0088] Compared with other feature point detection algorithms, the directed ring feature point algorithm can not only quickly detect and obtain the feature point information of the image to be stitched, but also retain the directionality of the feature point information, ensuring that the feature point directions of each image to be stitched are consistent, thereby improving the accuracy of subsequent feature point matching and thus improving the image stitching effect. By using a sliding window to perform feature point detection on each image to be stitched, the number of feature point detection times can be reduced and the efficiency of feature point detection can be improved.

[0089] Step S130 , based on the feature point information, a feature point matching algorithm is used to match the feature points of two adjacent images to be stitched, to obtain corresponding matching results.

[0090] Specifically, step S130 can take feature point information from two adjacent images to be stitched as input, repeatedly use the feature point matching algorithm, sequentially match each feature description matrix from one of the images to be stitched with each feature description matrix from the other image to be stitched, and use projection error to calculate whether a pair of feature description matrices match each other, until all feature description matrices included in the feature point information of the two adjacent images to be stitched are matched, and the corresponding matching results are output.

[0091] Exemplarily, the process of using the feature point matching algorithm to perform feature point matching in step S130 includes: for two adjacent images to be stitched, taking the feature description matrices corresponding to one of the images to be stitched as the target feature point information matrix in turn, and taking the feature description matrices corresponding to the other image to be stitched as the feature point information matrix to be matched; taking the difference between each target feature point information matrix and each feature point information matrix to be matched in turn, and calculating the projection difference value between the maximum gradient change contour in each target feature point information matrix and the maximum gradient change contour in each feature point information matrix to be matched; judging whether the minimum projection difference value corresponding to the target feature point information matrix is ​​greater than a preset threshold: if so, the matching result indicates that the feature point matching fails; if not, the matching result indicates that the feature point matching is successful.

[0092] Specific, combined Figure 4 , for two adjacent images to be stitched, assuming that a feature description matrix corresponding to one of the images to be stitched is Figure 4 The target feature point information matrix in the image to be stitched is a red-marked area with a maximum gradient change contour, and the feature description matrices corresponding to the other image to be stitched are respectively a to-be-matched feature point information matrix 1 to a to-be-matched feature point information matrix n, wherein the to-be-matched feature point information matrix 1 has a maximum gradient change contour as a blue-marked area, and the to-be-matched feature point information matrix n has a maximum gradient change contour as a red-marked area. When performing feature point matching, the target feature point information matrix is ​​subtracted from the to-be-matched feature point information matrix 1 to the to-be-matched feature point information matrix n one by one, and the maximum value contour difference is calculated. The projection difference values ​​of the maximum gradient change contour in the target feature point information matrix and the maximum gradient change contours in the to-be-matched feature point information matrix 1 to the to-be-matched feature point information matrix n are obtained, and then the minimum projection difference value corresponding to the target feature point information matrix is ​​found, and it is determined whether the minimum projection difference value found is greater than a preset threshold: if the minimum projection difference value found is greater than the preset threshold, it is determined that the feature point matching fails; if the minimum projection difference value found is less than or equal to the preset threshold, it is determined that the feature point matching succeeds. The value of the preset threshold can be set according to actual needs, and this embodiment does not limit this.

[0093] Compared with other feature point matching algorithms, this embodiment uses a simple difference calculation in the feature point matching algorithm, so that the determination of the overlapping area between two adjacent images to be stitched becomes simple, the calculation time is short, and the accuracy is high.

[0094] Step S140: If the matching result indicates that the feature points of the two adjacent images to be stitched are matched successfully, then the stitching amount corresponding to the two adjacent images to be stitched is determined according to the distance value of the successfully matched feature point pair.

[0095] Specifically, if the feature points of two adjacent images to be stitched are matched successfully, the successfully matched feature points can form a feature point pair, and the distance value of each feature point pair can be used as the stitching amount corresponding to the two adjacent images to be stitched. For example, the distance value of the feature point pair can be the maximum value, mean, median, mode, etc. of the distance between the two feature points in the feature point pair. The mode here refers to the value that appears the most times.

[0096] Exemplarily, in step S140, determining the stitching amount corresponding to the two adjacent images to be stitched according to the distance values ​​of the successfully matched feature point pairs includes: calculating the distance between the two feature points included in each feature point pair according to the absolute pixel coordinates of the midline points of each feature point on the respective images to be stitched, to obtain multiple feature point pair distances corresponding to the two adjacent images to be stitched; and taking the value that appears most frequently among the multiple feature point pair distances as the stitching amount corresponding to the two adjacent images to be stitched.

[0097] For example, assuming that the feature points A1, A2, ..., An of a certain image A to be stitched are successfully matched with the feature points B1, B2, ..., Bn of another adjacent image B to be stitched, then the pairs of successfully matched feature points are (A1, B1), (A2, B2), ..., (An, Bn). On this basis, according to the absolute pixel coordinates of the midline points of the feature points A1, A2, ..., An and B1, B2, ..., Bn on their respective images to be stitched, the distance A1B1 between the feature points A1 and B1, the distance A2B2 between the feature points A2 and B2, ..., and the distance AnBn between the feature points An and Bn are calculated respectively, and the value with the largest number of occurrences among A1B1, A2B2, ..., AnBn is taken as the stitching amount corresponding to the image A to be stitched and the image B to be stitched.

[0098] By taking the value that appears most frequently among the distances of multiple feature point pairs, i.e. the mode of the distances of the feature point pairs, as the stitching amount corresponding to the two adjacent images to be stitched, the matching rules of the feature point pairs can be better reflected, the image stitching error can be further reduced, and the image stitching effect can be effectively improved.

[0099] Step S150: If the matching result indicates that the feature points of the two adjacent images to be stitched fail to match, a global feature inheritance algorithm is used to determine the stitching amount corresponding to the two adjacent images to be stitched.

[0100] Specifically, when the feature points of two adjacent images to be stitched fail to match, the distance between the feature point pairs cannot be directly calculated. Therefore, step S150 uses a global feature inheritance algorithm to determine the stitching amount corresponding to the two adjacent images to be stitched.

[0101] Exemplarily, in step S150, a global feature inheritance algorithm is used to determine the stitching amount corresponding to the two adjacent images to be stitched, including: based on the stitching position of the two adjacent images to be stitched where feature point matching fails, searching around for other two adjacent images to be stitched that have the same stitching direction as the two adjacent images to be stitched and have successful feature point matching; and determining the stitching amount corresponding to the two adjacent images to be stitched where feature point matching fails based on feature point information corresponding to the other two adjacent images to be stitched that are searched.

[0102] Specifically, when determining the stitching amount according to the feature point information, step S150 may adopt a method similar to step S140, that is, determining the corresponding stitching amount according to the distance value of the successfully matched feature point pair, preferably the distance mode.

[0103] Compared with the hard stitching method adopted by other algorithms when feature points cannot be successfully matched, the global feature inheritance algorithm provided by this embodiment can obtain feature point information from the other two adjacent images to be stitched that have successfully matched feature points when the feature points of the two adjacent images to be stitched fail to match or when there are no feature points, and determine the approximate stitching amount based on the feature point information, thereby minimizing the error to the greatest extent. At the same time, since the stitching amount determined based on the feature point information corresponding to the other two adjacent images to be stitched that have successfully matched feature points actually compensates for the deviation caused by the shaking of the camera, the global feature inheritance algorithm provided by this embodiment can also compensate for the deviation caused by the shaking of the camera, greatly improving the image stitching quality.

[0104] Step S160 , stitching and fusing the images to be stitched according to the stitching amount to obtain a target stitching image.

[0105] Specifically, step S160 may use an image fusion algorithm to fuse each of the two adjacent images to be stitched in turn according to the stitching amounts corresponding to each of the two adjacent images to be stitched, so as to achieve image stitching. The image fusion algorithm may include but is not limited to a weighted fusion algorithm, a maximum fusion algorithm, a minimum fusion algorithm, an average fusion algorithm, etc.

[0106] Exemplarily, step S160 preferably uses a weighted fusion algorithm to achieve image stitching to improve the quality of image stitching. At this time, step S160 may include: determining the image overlap area corresponding to each of the two adjacent images to be stitched according to the stitching amount corresponding to each of the two adjacent images to be stitched; determining the weight by a logarithmic function, i.e., a log function, and superimposing and fusing the pixels in the image overlap area corresponding to each of the two adjacent images to be stitched according to the weight to obtain a target stitched image.

[0107] For example, assuming that the stitching amount corresponding to the image 1 to be stitched and the image 2 to be stitched on its right side is 1836.20 pixels, then in step S160, the area located on the right side of the image 1 to be stitched and having an area of ​​1836.20 pixels can be used as the image overlapping area in the image 1 to be stitched, and the area located on the left side of the image 2 to be stitched and having an area of ​​1836.20 pixels can be used as the image overlapping area in the image 2 to be stitched. The pixel weight of the image overlapping area is determined by the log function, and the image overlapping area in the image 1 to be stitched and the image overlapping area in the image 2 to be stitched are superimposed and fused according to the pixel weight. When all two adjacent images to be stitched are stitched and fused one by one, the target stitched image can be obtained.

[0108] It should be noted that when using the log function to determine the weight, the weight of each pixel in the overlapping area of ​​the image can be made to satisfy the requirement that the closer to the fusion center position, the larger the value, thereby ensuring a natural gradient of image color in the overlapping area of ​​the image and improving the image stitching quality.

[0109] Compared with the prior art, the image stitching method provided by the embodiment of the present disclosure uses the images to be stitched taken according to the preset overlap ratio for stitching, which can ensure the stability of the approximate range of the stitching amount and improve the stability of the image stitching. In view of the difficulty of uncertain errors in the image stitching process, a feature point algorithm is designed, and a feature point matching algorithm is used to determine the feature point pairs of two adjacent images to be stitched, and then the stitching amount of the two adjacent images to be stitched is determined according to the distance value of the feature point pairs. It can not only quickly detect the feature point information of the images to be stitched, but also retain the directionality of the feature point information, improve the accuracy of feature point matching, and reduce the time-consuming feature matching calculation. In view of the situation that there may be no obvious target or most of the blanks in the image to be stitched, which leads to the failure of feature detection and matching, a global feature inheritance algorithm is designed to realize the stitching compensation of the feature detection and matching failure. By stitching and fusing each image to be stitched according to the stitching amount, the image stitching quality is also improved.

[0110] In order to enable those skilled in the art to better understand the above implementation, a specific example is provided below for illustration.

[0111] For example, Figure 5As shown, a microscopic image stitching method includes: first, using a high-content instrument to shoot a group of microscopic images according to a preset overlap ratio to obtain images 1 to 9 with overlapping areas. Then, the directed ring feature point algorithm is used to extract feature points from two adjacent images in sequence, and the feature points of the extracted two adjacent images are matched using a feature point matching algorithm. The mode of the distance between the feature point pairs that successfully match between the two adjacent images is used as the stitching amount of the two adjacent images. The global feature inheritance algorithm is applied to the two adjacent images where the feature point matching fails, and the stitching amount of the two adjacent images where the feature point matching fails is determined based on the feature point information of other two adjacent images with the same stitching direction and successful feature point matching. Finally, according to the stitching amount of the two adjacent images, a weighted fusion algorithm is applied to fuse each image with its adjacent image in sequence, and finally the microscopic image stitching is realized to obtain a stitching result image.

[0112] For example, Figure 6 As shown, an image stitching method comprises the following steps:

[0113] Step S10, photographing the images to be stitched: using a high-content instrument, according to the input of the puzzle, photograph a group of images to be stitched at a preset overlap ratio.

[0114] Step S11, judging whether all images have been stitched, that is, judging whether all images to be stitched have been stitched: if so, that is, all stitching has been completed, then the process ends; if not, that is, all stitching has not been completed, then the process proceeds to step S12.

[0115] Step S12, using a directed ring feature point algorithm to detect feature points of adjacent images to be spliced: using a directed ring feature point algorithm, detect two adjacent images to be spliced, calculate the feature points in the two images to be spliced, and obtain corresponding feature point information. The feature point information corresponding to each image to be spliced ​​usually includes information of multiple feature points, and the information of each feature point includes the center position of the feature point, i.e., the feature point coordinates, and the feature description matrix of the feature point.

[0116] Step S13, using a feature point matching algorithm to match feature points of adjacent images: repeatedly using the feature point matching algorithm, taking the feature description matrix of each adjacent image to be stitched as input in turn, using the projection error to calculate whether the feature points of the two adjacent images to be stitched match each other, until all feature points of the two adjacent images to be stitched are matched, and outputting a matching result indicating whether the match is successful.

[0117] Step S14, judging whether the feature points are matched successfully: judging whether the two adjacent images to be stitched are matched successfully, if they are matched successfully, then entering the step of determining the stitching amount according to the successfully matched feature points (including step S15 and step S16), if the match fails, then entering the step of determining the stitching amount using the global feature inheritance algorithm, i.e., step S17.

[0118] Step S15, calculating the distances of all feature point pairs: based on all feature point pairs that are successfully matched between two adjacent images to be stitched, the distance between two feature points included in each feature point pair is calculated according to the absolute pixel coordinates of the midline points of the feature points on the respective images to be stitched.

[0119] Step S16, determining the stitching amount according to the mode of the feature point pair distances: according to the distances between the two feature points corresponding to all the feature point pairs of the two adjacent images to be stitched calculated in step S15, counting the number of times each distance appears, and selecting the distance with the largest number of length appearances as the stitching amount of the two adjacent images to be stitched.

[0120] Step S17, using a global feature inheritance algorithm to determine the stitching amount: based on the stitching position of the two adjacent images to be stitched where the feature point matching fails, search for other two adjacent images to be stitched that have the same stitching direction and have successful feature point matching nearby, inherit the feature point information of the other two adjacent images to be stitched, and determine the stitching amount of the two adjacent images to be stitched where the feature point matching fails based on the inherited feature point information.

[0121] Step S18, using a weighted fusion algorithm to fuse the images according to the stitching amount: determine the size of the corresponding image overlap area according to the stitching amount of each two adjacent images to be stitched, determine the weight according to the size of the image overlap area using a log function, and superimpose and fuse the two adjacent images to be stitched in turn according to the weight, and return to step S11.

[0122] Another embodiment of the present disclosure relates to a method for stitching images of an in vitro three-dimensional biological model, which is applied to in vitro three-dimensional biological model imaging images and includes the image stitching method described in the above embodiment.

[0123] Specifically, by setting the image to be stitched in the image stitching method provided in the above embodiment as an in vitro three-dimensional biological model imaging image, the stitching method of the in vitro three-dimensional biological model imaging image provided in this embodiment can be obtained, thereby effectively reducing the influence of random errors on the stitching process of the in vitro three-dimensional biological model imaging image, and improving the stitching stability and stitching quality of the in vitro three-dimensional biological model imaging image.

[0124] Another embodiment of the present disclosure relates to an image stitching device, such as Figure 7 As shown, including:

[0125] An acquisition module 710 is used to acquire a set of images to be stitched that are taken according to a preset overlap ratio;

[0126] The detection module 720 is used to detect feature points of each image to be spliced ​​using a feature point algorithm to obtain feature point information corresponding to each image to be spliced;

[0127] A matching module 730 is used to match feature points of two adjacent images to be stitched based on feature point information using a feature point matching algorithm to obtain corresponding matching results;

[0128] The determination module 740 is used to: if the matching result indicates that the feature points of the two adjacent images to be stitched are matched successfully, determine the stitching amount corresponding to the two adjacent images to be stitched according to the distance value of the successfully matched feature point pair; if the matching result indicates that the feature points of the two adjacent images to be stitched fail to match, determine the stitching amount corresponding to the two adjacent images to be stitched using a global feature inheritance algorithm;

[0129] The stitching and fusion module 750 is used to stitch and fuse the images to be stitched according to the stitching amount to obtain a target stitching image.

[0130] Exemplarily, the feature point information includes feature point coordinates and a feature description matrix; and the feature point algorithm includes a directed ring feature point algorithm.

[0131] The detection module 720 is used to detect feature points of each image to be spliced ​​using a feature point algorithm to obtain feature point information corresponding to each image to be spliced, including:

[0132] The detection module 720 is used to: use a circular sliding window with a preset diameter to perform feature point detection on each image to be stitched, use the center coordinates of the circular sliding window as the feature point coordinates, calculate the change gradients of all pixels in the circular sliding window and their adjacent pixels, generate a corresponding gradient size matrix, mark the contour with the maximum gradient change in the gradient size matrix, and obtain a feature description matrix.

[0133] Exemplarily, the matching module 730 is used to match the feature points of two adjacent images to be stitched using a feature point matching algorithm based on the feature point information to obtain corresponding matching results, including:

[0134] The matching module 730 is used to:

[0135] For two adjacent images to be stitched, each feature description matrix corresponding to one of the images to be stitched is used as the target feature point information matrix in turn, and each feature description matrix corresponding to the other image to be stitched is used as the feature point information matrix to be matched;

[0136] Subtract each target feature point information matrix from each feature point information matrix to be matched one by one, and calculate the projection difference value between the maximum gradient change contour in each target feature point information matrix and the maximum gradient change contour in each feature point information matrix to be matched;

[0137] Determine whether the minimum projection difference value corresponding to the target feature point information matrix is ​​greater than a preset threshold: if so, the matching result indicates that the feature point matching fails; if not, the matching result indicates that the feature point matching succeeds.

[0138] Exemplarily, the determination module 740 is used to determine the stitching amount corresponding to the two adjacent images to be stitched according to the distance value of the successfully matched feature point pair, including:

[0139] The determination module 740 is used to:

[0140] According to the absolute pixel coordinates of the midline points of each feature point on the respective images to be stitched, the distance between the two feature points included in each feature point pair is calculated respectively, so as to obtain the distances of multiple feature point pairs corresponding to two adjacent images to be stitched;

[0141] The value that appears most frequently among the distances between the multiple feature point pairs is used as the stitching amount corresponding to the two adjacent images to be stitched.

[0142] Exemplarily, the determination module 740 is used to determine the stitching amount corresponding to the two adjacent images to be stitched by using a global feature inheritance algorithm, including:

[0143] The determination module 740 is used to:

[0144] Based on the stitching positions of the two adjacent images to be stitched where the feature point matching fails, searching around for other two adjacent images to be stitched that have the same stitching direction as the two adjacent images to be stitched and whose feature points match successfully;

[0145] The stitching amount corresponding to the two adjacent images to be stitched where the feature point matching fails is determined according to the stitching amount corresponding to the other two adjacent images to be stitched that are searched.

[0146] Exemplarily, the stitching and fusion module 750 is used to stitch and fuse the images to be stitched according to the stitching amount to obtain a target stitched image, including:

[0147] The splicing and fusion module 750 is used for:

[0148] Determine the image overlap area corresponding to each of the two adjacent images to be stitched according to the stitching amount corresponding to each of the two adjacent images to be stitched;

[0149] The weights are determined by a logarithmic function, and the pixels in the overlapping area of ​​each adjacent image to be stitched are superimposed and fused according to the weights to obtain a target stitched image.

[0150] Exemplarily, the number b of images to be stitched satisfies: b≥(ad+x) / (dx); wherein a represents the number of images to be stitched, d represents the field of view size of a single image to be stitched, and x represents a preset overlap ratio.

[0151] The specific implementation method of the image stitching device provided in the embodiment of the present disclosure can refer to the image stitching method provided in the embodiment of the present disclosure, which will not be repeated here.

[0152] The image stitching device provided by the embodiment of the present disclosure, compared with the prior art, uses the acquisition module to stitch the images to be stitched according to the preset overlap ratio, which can ensure the stability of the approximate range of the stitching amount and improve the stability of the image stitching. In view of the difficulty of uncertain errors in the image stitching process, the detection module designs a feature point algorithm, and uses the feature point matching algorithm in the matching module to determine the feature point pairs of two adjacent images to be stitched, and then determines the stitching amount of two adjacent images to be stitched according to the distance value of the feature point pairs in the determination module, which can not only quickly detect the feature point information of the images to be stitched, but also retain the directionality of the feature point information, improve the accuracy of feature point matching, and reduce the time-consuming feature matching calculation. In view of the situation that there may be no obvious target or most of the blanks in the image to be stitched, resulting in the failure of feature detection matching, the determination module also designs a global feature inheritance algorithm to realize the stitching compensation for the failure of feature detection matching. Through the stitching and fusion module, each image to be stitched is stitched and fused according to the stitching amount, and the image stitching quality is also improved.

[0153] Another embodiment of the present disclosure relates to an electronic device, such as Figure 8 As shown, including:

[0154] at least one processor 801; and,

[0155] A memory 802 is communicatively connected to at least one processor 801; wherein,

[0156] The memory 802 stores instructions that can be executed by at least one processor 801, and the instructions are executed by at least one processor 801 so that the at least one processor 801 can execute the image stitching method described in the above embodiment or the stitching method of in vitro three-dimensional biological model imaging images.

[0157] Among them, the memory and the processor are connected in a bus manner, and the bus may include any number of interconnected buses and bridges, and the bus connects various circuits of one or more processors and memories together. The bus can also connect various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. The bus interface provides an interface between the bus and the transceiver. The transceiver can be one element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices on a transmission medium. The data processed by the processor is transmitted on a wireless medium via an antenna, and further, the antenna also receives data and transmits the data to the processor.

[0158] The processor is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory can be used to store data used by the processor when performing operations.

[0159] Another embodiment of the present disclosure relates to a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the image stitching method described in the above embodiment or the image stitching method of an in vitro three-dimensional biological model imaging.

[0160] That is, those skilled in the art can understand that all or part of the steps in the method described in the above embodiments can be completed by instructing the relevant hardware through a program, and the program is stored in a storage medium, including a number of instructions to enable a device (which can be a single-chip microcomputer, chip, etc.) or a processor to execute all or part of the steps of the method described in each embodiment of the present disclosure. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store program codes.

[0161] Those skilled in the art will appreciate that the above-mentioned embodiments are specific embodiments for implementing the present disclosure, and in actual applications, various changes may be made thereto in form and detail without departing from the spirit and scope of the present disclosure.

Claims

1. An image stitching method, characterized in that: The image stitching method comprises: Acquire a set of images to be stitched that are taken according to a preset overlap ratio; Using a feature point algorithm to detect feature points of each of the images to be spliced, to obtain feature point information corresponding to each of the images to be spliced; Based on the feature point information, feature points of two adjacent images to be stitched are matched using a feature point matching algorithm to obtain corresponding matching results; If the matching result indicates that the feature points of the two adjacent images to be stitched are matched successfully, then determining the stitching amount corresponding to the two adjacent images to be stitched according to the distance value of the successfully matched feature point pair; If the matching result indicates that the feature points of the two adjacent images to be stitched fail to match, a global feature inheritance algorithm is used to determine the stitching amount corresponding to the two adjacent images to be stitched; According to the stitching amount, each of the images to be stitched is stitched and merged to obtain a target stitching image; The method of using a global feature inheritance algorithm to determine the stitching amount corresponding to the two adjacent images to be stitched includes: Based on the stitching position of the two adjacent images to be stitched where the feature point matching fails, searching around for other two adjacent images to be stitched that have the same stitching direction as the two adjacent images to be stitched and have successful feature point matching; Determine the stitching amount corresponding to the two adjacent images to be stitched whose feature points fail to match, according to the feature point information corresponding to the other two adjacent images to be stitched that are searched; The feature point information includes a feature description matrix; The method of matching the feature points of two adjacent images to be stitched using a feature point matching algorithm based on the feature point information to obtain a corresponding matching result includes: For two adjacent images to be stitched, the feature description matrices corresponding to one of the images to be stitched are sequentially used as target feature point information matrices, and the feature description matrices corresponding to the other image to be stitched are respectively used as feature point information matrices to be matched; Subtracting each of the target feature point information matrices from each of the feature point information matrices to be matched one by one in turn, and calculating the projection difference value between the maximum gradient change contour in each of the target feature point information matrices and the maximum gradient change contour in each of the feature point information matrices to be matched; Determine whether the minimum projection difference value corresponding to the target feature point information matrix is ​​greater than a preset threshold: if so, the matching result indicates that the feature point matching fails; if not, the matching result indicates that the feature point matching succeeds.

2. The image stitching method according to claim 1, characterized in that: The feature point information also includes feature point coordinates; the feature point algorithm includes a directed ring feature point algorithm; The step of using a feature point algorithm to detect feature points of each of the images to be stitched to obtain feature point information corresponding to each of the images to be stitched includes: A circular sliding window with a preset diameter is used to perform feature point detection on each of the images to be stitched, and the center coordinates of the circular sliding window are used as the feature point coordinates. The change gradients of all pixel points in the circular sliding window and their adjacent pixel points are calculated to generate a corresponding gradient size matrix. The contour with the maximum gradient change in the gradient size matrix is ​​marked to obtain the feature description matrix.

3. The image stitching method according to claim 1 or 2, characterized in that: The step of determining the stitching amount corresponding to the two adjacent images to be stitched according to the distance values ​​of the successfully matched feature point pairs includes: According to the absolute pixel coordinates of the midline points of each feature point on the images to be stitched, respectively calculating the distance between the two feature points included in each feature point pair, to obtain a plurality of feature point pair distances corresponding to two adjacent images to be stitched; The value with the largest number of occurrences among the distances between the plurality of feature point pairs is used as the stitching amount corresponding to the two adjacent images to be stitched.

4. The image stitching method according to claim 1 or 2, characterized in that: The step of performing stitching and fusing the images to be stitched according to the stitching amount to obtain a target stitching image includes: Determining, according to the stitching amounts corresponding to the two adjacent images to be stitched, image overlapping areas corresponding to the two adjacent images to be stitched; The weights are determined by a logarithmic function, and the pixels in the image overlap region corresponding to two adjacent images to be stitched are superimposed and fused according to the weights to obtain the target stitched image.

5. The image stitching method according to claim 1 or 2, characterized in that: The number b of the images to be stitched satisfies: b≥(ad+x) / (dx); wherein a represents the number of images to be stitched, d represents the field of view size of a single image to be stitched, and x represents the preset overlap ratio.

6. A method for stitching images of an in vitro three-dimensional biological model, characterized in that: Applicable to in vitro three-dimensional biological model imaging images, including the image stitching method described in any one of claims 1 to 5.

7. An image stitching device, characterized in that: The image stitching device comprises: An acquisition module, used for acquiring a set of images to be stitched that are shot according to a preset overlap ratio; A detection module, used to perform feature point detection on each of the images to be spliced ​​using a feature point algorithm to obtain feature point information corresponding to each of the images to be spliced; A matching module, used to match the feature points of two adjacent images to be stitched using a feature point matching algorithm based on the feature point information to obtain a corresponding matching result; A determination module, configured to: if the matching result indicates that the feature points of the two adjacent images to be stitched are matched successfully, determine the stitching amount corresponding to the two adjacent images to be stitched according to the distance value of the successfully matched feature point pair; if the matching result indicates that the feature points of the two adjacent images to be stitched fail to match, determine the stitching amount corresponding to the two adjacent images to be stitched using a global feature inheritance algorithm; A stitching and fusion module, used for stitching and fusion of the images to be stitched according to the stitching amount, to obtain a target stitching image; The method of using a global feature inheritance algorithm to determine the stitching amount corresponding to the two adjacent images to be stitched includes: Based on the stitching position of the two adjacent images to be stitched where the feature point matching fails, searching around for other two adjacent images to be stitched that have the same stitching direction as the two adjacent images to be stitched and have successful feature point matching; Determine the stitching amount corresponding to the two adjacent images to be stitched whose feature points fail to match, according to the feature point information corresponding to the other two adjacent images to be stitched that are searched; The feature point information includes a feature description matrix; The method of matching the feature points of two adjacent images to be stitched using a feature point matching algorithm based on the feature point information to obtain a corresponding matching result includes: For two adjacent images to be stitched, the feature description matrices corresponding to one of the images to be stitched are sequentially used as target feature point information matrices, and the feature description matrices corresponding to the other image to be stitched are respectively used as feature point information matrices to be matched; Subtracting each of the target feature point information matrices from each of the feature point information matrices to be matched one by one in turn, and calculating the projection difference value between the maximum gradient change contour in each of the target feature point information matrices and the maximum gradient change contour in each of the feature point information matrices to be matched; Determine whether the minimum projection difference value corresponding to the target feature point information matrix is ​​greater than a preset threshold: if so, the matching result indicates that the feature point matching fails; if not, the matching result indicates that the feature point matching succeeds.

8. An electronic device, characterized in that: include: at least one processor; as well as, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the image stitching method described in any one of claims 1 to 5 or the stitching method of in vitro three-dimensional biological model imaging images described in claim 6.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the image stitching method described in any one of claims 1 to 5 or the image stitching method of an in vitro three-dimensional biological model imaging described in claim 6 is implemented.

Citation Information

Patent Citations

  • Tunnel distribution map splicing parameter processing method based on Sift characteristic

    CN108109112A

  • Microscopic image splicing method and system

    CN113205457A

  • Image splicing method and device, equipment and storage medium

    CN113724135A

  • Image splicing method, electronic equipment and computer readable storage medium

    CN114418839A

  • Rapid wide-angle stitching method for high-resolution images

    WO2021169334A1