Skin CT Large Field-of-View Image Mosaic Method and System with High Reliability and Fault Tolerance

Abstract

The present invention discloses a method and system for stitching large field-of-view images of a dermoscopy confocal microscope with high reliability and fault tolerance. The method includes: 1) Manually determining the starting point and size of the stitching area, and setting M rows × N columns of dermoscopy confocal microscope images to be stitched in a grid shape, ensuring that there are overlapping areas between two adjacent images on the left and right, as well as above and below; 2) Making the objective lens of the dermoscope start from the starting point of the stitching area and scan each grid in the first row from left to right in sequence, then switch to the next row and move in the reverse direction, and so on, completing the scanning of all grids along an S-shaped trajectory, and performing stitching while acquiring images. By introducing a fault tolerance mechanism, the present invention can ensure that during the large field-of-view stitching process, the overall stitching will not fail due to local image stitching errors, but rather limit the error area to a small local range, thereby improving the robustness and efficiency of large field-of-view image stitching.

Description

Technical Field

[0001] The present invention relates to the field of skin CT detection, and particularly to a method and system for stitching large field-of-view skin CT images with high reliability and fault tolerance. Background Art

[0002] Skin CT imaging is based on the confocal principle. An in-vivo scanning device is added to a confocal laser scanning microscope, thereby realizing non-invasive observation of cell characteristics at different structural levels of the skin.

[0003] Skin CT obtains grayscale images by scanning subcutaneous tissues. These images are usually grayscale images with varying degrees of brightness based on the different refractive indices of microstructures in the skin tissue, such as melanin, oxyhemoglobin, and cell organelles. During the scanning process of skin CT, if the laser is too strong (too weak), the overall image will be too bright (dark), resulting in incorrect differentiation of different tissue structures. Therefore, it is necessary to adjust the laser intensity for different subcutaneous depths to obtain high-quality skin CT images with appropriate brightness.

[0004] Skin CT improves resolution by sacrificing the field of view and usually captures local areas. When using skin CT for diagnosis, not only local observation is required, but also comprehensive detection and analysis are needed. Therefore, in order to obtain large field-of-view skin CT images, multiple skin CT images need to be stitched together.

[0005] The existing image stitching algorithms mainly include the following several types:

[0006] One is to find the displacement vector of the images to be stitched based on image feature point matching. By extracting the feature points in the overlapping area of two images and performing matching, the overall displacement of the images is calculated. Commonly used image feature operators include HARRIS, SURF, SIFT, etc., which have good effects on natural images with strong features. However, skin CT images have characteristics such as high noise and unstable feature details. Therefore, for skin CT images, there are difficulties such as difficult extraction of feature points and large matching errors in the feature point matching method.

[0007] Two is to find the displacement vector of the images to be stitched based on the phase correlation method. By calculating the displacement vector of two images to be matched in the frequency domain, image stitching is realized in the spatial domain. This method is prone to pixel-level errors, and when the magnification of the microscope image is relatively high, the errors are easily amplified, affecting the stitching visual effect.

[0008] Three is to calculate the correlation degree of the overlapping area under different displacement vectors by traversing the displacement vectors, and use the displacement vector with the highest correlation degree as the displacement vector of the images to be stitched for stitching. Although this method is applicable to the stitching of skin CT images, its computational complexity is too high, seriously affecting the real-time performance of stitching.

[0009] The skin CT is a device for real-time detection of the human skin. Its imaging is vulnerable to external jitter and internal skin tissue activities, resulting in certain changes in the CT images of the same area at different times. In addition, during the stitching process, the microscope objective needs to be moved. Usually, a motor drives the objective to move and take pictures, and the motor will also have a certain deviation during the process of moving a fixed number of steps, resulting in inaccurate stitching of the images.

[0010] Therefore, a more reliable solution is needed now. Summary of the Invention

[0011] The technical problem to be solved by the present invention is to provide a large-field skin CT image stitching method and system with high reliability and fault tolerance in view of the deficiencies in the above-mentioned prior art. By introducing a fault tolerance mechanism, the present invention can ensure that the overall stitching will not fail due to local image stitching errors during the large-field stitching process, but limit the error area to a small local range, thereby improving the robustness of the large-field image stitching.

[0012] To achieve the above object, the technical solution adopted by the present invention is: A large-field skin CT image stitching method with high reliability and fault tolerance, including:

[0013] 1) Manually determine the starting point and size of the stitching area, and set M rows × N columns of skin CT images to be stitched in a grid shape, ensuring that there is an overlapping area between two adjacent images on the left and right, and above and below.

[0014] 2) Make the objective of the dermoscope start from the starting point of the stitching area and scan each grid in the first row from left to right in turn, then switch to the next row and move in the reverse direction, and so on, and complete the scanning of all grids along the S-shaped trajectory.

[0015] During the process of the objective of the dermoscope scanning, first obtain the skin CT image of the current grid, and then stitch the obtained skin CT image onto the large-field skin CT image that has been stitched before scanning the current grid, so as to stitch while obtaining images until the acquisition and stitching of M rows × N columns of skin CT images are completed, and a complete large-field skin CT image is obtained.

[0016] Preferably, in step 2), after obtaining the skin CT image of the current grid, first detect whether the brightness of the current image is within the normal range, and thereby judge whether it is necessary to adjust the incident laser intensity of the dermoscope to ensure that an image with a satisfactory brightness is obtained;

[0017] Among them, the brightness of the image is judged whether it is within the normal range through the image gray mean value mean and the average deviation mdev. The formula is:

[0018] mean = ∑(x i -ref) / N;

[0019] mdev = ∑|(i - ref) - mean|×His[i] / ∑His[i];

[0020] where the average grayscale value mean of the image is the average value after the image is subtracted by the reference value, x i is the grayscale value of the current image pixel, ref is the set reference value, N is the number of image pixels; His[i] represents the grayscale histogram of the image, and the range of i is 0 - 255;

[0021] Set the deviation threshold T according to the skin CT image features m , if mdev is less than T, it means the brightness is normal; if mdev is less than T, it means the brightness is abnormal, and judge whether the image is too bright or too dark according to the mean value: if mean is greater than 0, it means the image is too bright and the laser intensity needs to be reduced, if mean is less than 0, it means the image is too dark and the laser intensity needs to be increased; after adjusting the laser, retake the skin CT image of the current grid and calculate and judge whether the image brightness is normal until an image with satisfactory brightness is obtained, and then start splicing.

[0022] Preferably, assign a splicing flag bit to each image. If the splicing of the image is successful, it is marked as correct, and if the splicing fails, it is marked as incorrect;

[0023] The method for judging whether the image splicing is successful or not is as follows:

[0024] When splicing an image with an incorrect splicing flag bit, the image to be spliced that is spliced with the image to be spliced is directly marked as incorrect;

[0025] When splicing an image with a correct splicing flag bit, calculate the maximum value Max of the matching degree Ncc of the overlapping area of the two images after splicing the image to be spliced with the image to be spliced Ncc with the set matching degree threshold T N in size. If Max Ncc > T N , then mark the image to be spliced as correct, otherwise mark the image to be spliced as incorrect.

[0026] Preferably, the calculation formula for the matching degree Ncc of two images in the overlapping area is:

[0027]

[0028] where f represents the pre - matching image pixel value, t represents the template image pixel value, μ represents the pixel average value, σ represents the standard deviation, and n represents the total number of pixels.

[0029] Preferably, the method for stitching the first row of images, the first column of images, and the last column of images is as follows: when stitching the first row of images, stitch the current image with the previous image in the moving direction; when stitching the first column or the last column of images, stitch the current image with the image in the same column as the current image and in the row above the current image;

[0030] Among them, if the stitching flag bit of the image to be stitched is incorrect, directly mark the current image as incorrect;

[0031] If the stitching flag bit of the image to be stitched is correct, then calculate the maximum value Max of the matching degree Ncc of the overlapping area of the two images after stitching Ncc , if Max Ncc > T N , then mark the current image as correct, otherwise mark the current image as incorrect.

[0032] Preferably, the method for stitching the images at the remaining positions except for the first row of images, the first column of images, and the last column of images is as follows:

[0033] Denote the current image as P0, the previous image in the moving direction of the current image as P1, and the image in the same column as the previous image and in the row above the current image as P2;

[0034] If one of P1 and P2 is marked as correct and the other is marked as incorrect, then P0 selects the image marked as correct for stitching, and then calculates the maximum value Max of the matching degree Ncc of the overlapping area of the two images after stitching Ncc , if Max Ncc > T N , then mark the current image as correct, otherwise mark the current image as incorrect;

[0035] If both P1 and P2 are marked as correct, then calculate the maximum value Ncc01 of the matching degree of the overlapping area after stitching P0 and P1 and the maximum value Ncc02 of the matching degree of the overlapping area after stitching P0 and P2. When Ncc01 is greater than Ncc02, P0 selects to stitch with P1, and when Ncc01 > T N At this time, mark the current image as correct, otherwise mark the current image as incorrect; when Ncc01 is not greater than Ncc02, P0 selects to stitch with P2, and when Ncc02 > T N At this time, mark the current image as correct, otherwise mark the current image as incorrect;

[0036] If both P1 and P2 are marked as incorrect, then P0 is stitched according to the preset offset, and the current image is marked as incorrect.

[0037] Preferably, when splicing, calculate the optimal displacement vector of the current image, and then move the current image according to the optimal displacement vector for splicing;

[0038] For the spliced image, use the gray-scale weighted average fusion algorithm to perform weighted fusion on the overlapping area of the two spliced images, and use the pixel value result after weighted fusion as the pixel value of the overlapping area. The formula of the weighted average fusion algorithm is as follows:

[0039] V = a i ×V 1 +(1 - a i )×V 2 , a i = w - i / w;

[0040] Wherein, V is the pixel value after fusion, V 1 , V 2 respectively represent the pixel values of the image to be spliced and the image to be spliced, a i represents the weight value, w represents the fusion width, and i represents the abscissa of the current pixel point.

[0041] The present invention also provides a high-reliability and fault-tolerant skin CT large-field-of-view image splicing system, which uses the above method to splice skin CT large-field-of-view images.

[0042] The present invention also provides a storage medium, on which a computer program is stored, and is characterized in that the program is used to implement the above method when executed.

[0043] The present invention also provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, and is characterized in that the processor implements the above method when executing the computer program.

[0044] The beneficial effects of the present invention are:

[0045] The high-reliability and fault-tolerant skin CT large-field-of-view image splicing method provided by the present invention realizes limiting the influence of partial image splicing errors caused by jitter and other reasons to a small area during the large-field-of-view skin CT splicing process through the introduction of a fault-tolerant mechanism, and can ensure the integrity and robustness of the large-field-of-view splicing;

[0046] The present invention greatly improves the calculation speed of traversal search and significantly improves the calculation efficiency of the algorithm by introducing a GPU acceleration mechanism, which facilitates the edge scanning and splicing of skin CT images, helps to reduce the patient jitter during skin CT image acquisition, and improves the splicing accuracy. Brief Description of the Drawings

[0047] Figure 1Flow chart of the large field of view image stitching method for skin CT with high reliability and fault tolerance according to the present invention;

[0048] Figure 2 Schematic diagram of the stitching movement strategy in Embodiment 1 of the present invention;

[0049] Figure 3 Schematic diagram of stitching in Embodiment 1 of the present invention;

[0050] Figures 4(a)-(f) are diagrams showing the stitching process of 4×4 sized skin CT images in an embodiment. Detailed implementation manners

[0051] The following further elaborates on the present invention in conjunction with embodiments, so that those skilled in the art can implement it with reference to the text of the specification.

[0052] It should be understood that terms such as "having", "comprising", and "including" as used herein do not exclude the presence or addition of one or more other elements or combinations thereof.

[0053] Embodiment 1

[0054] Refer to Figure 1 , a large field of view image stitching method for skin CT with high reliability and fault tolerance in this embodiment includes:

[0055] I. Manually determine the starting point and size of the stitching area, and set M rows × N columns of skin CT images to be stitched in the shape of a grid, ensuring that there is an overlapping area between two adjacent images on the left and right, and above and below.

[0056] II. Make the objective lens of the dermoscope start from the starting point of the stitching area and scan each grid in the first row from left to right in sequence, then switch to the next row and move in the reverse direction, and so on in a cycle, and complete the scanning of all grids along the S-shaped trajectory, refer to Figure 2 ;

[0057] During the process of the objective lens of the dermoscope scanning, first obtain the skin CT image of the current grid, and then stitch the obtained skin CT image onto the large field of view skin CT image that has been stitched before scanning the current grid, so as to stitch while obtaining images until the acquisition and stitching of M rows × N columns of skin CT images are completed, and a complete large field of view skin CT image is obtained.

[0058] In step 2), after obtaining the skin CT image of the current grid, first detect whether the brightness of the current image is within the normal range, and thereby judge whether it is necessary to adjust the incident laser intensity of the dermoscope to ensure that an image with satisfactory brightness is obtained;

[0059] Among them, the mean and mean deviation mdev of the image grayscale are used to determine whether the brightness of the image is within the normal range. When there is abnormal brightness, the mean will deviate from the mean point (which can be assumed to be 128), and the mean deviation will also be small. The specific formula is as follows:

[0060] mean = ∑(x i - ref) / N;

[0061] mdev = ∑|(i - ref)-mean|×His[i] / ∑His[i];

[0062] Among them, the mean of the image grayscale mean is the mean after the image subtracts the reference value. x i is the grayscale value of the current image pixel, ref is the set reference value (usually 128), N is the number of image pixels; His[i] represents the image grayscale histogram, and the range of i is 0 - 255;

[0063] Set the deviation threshold T m according to the skin CT image features. If mdev is less than T, it means the brightness is normal; if mdev is less than T, it means the brightness is abnormal, and judge whether the image is too bright or too dark according to the mean value: if mean is greater than 0, it means the image is too bright and the laser intensity needs to be reduced; if mean is less than 0, it means the image is too dark and the laser intensity needs to be increased; After adjusting the laser, retake the skin CT image of the current grid and calculate and judge whether the image brightness is normal until an image with satisfactory brightness is obtained, and then start stitching.

[0064] The steps for formal stitching include:

[0065] 1. Preprocess the original skin CT image obtained by shooting. Since the skin CT image has large imaging noise, it needs to be smoothed by Gaussian filtering; in addition, when the skin CT image is taken, the laser is adjusted based on the overall grayscale value of the image, and there may be local over - brightness or over - darkness in the image. Therefore, the image is enhanced by the adaptive contrast - limited histogram equalization method to highlight the features of the image;

[0066] 2. Traverse and match the pre - processed images to be stitched to achieve stitching.

[0067] 2 - 1. Here, a fault - tolerance mechanism is introduced. First, assign a stitching flag bit to each image. If the stitching of the image is successful, it is marked as correct; if the stitching fails, it is marked as wrong;

[0068] The method for judging whether the image stitching is successful or failed is:

[0069] When stitching an image with a wrong stitching flag bit, the image to be stitched that is stitched with the image to be stitched is directly marked as wrong;

[0070] When stitching an image with a correct stitching flag bit, calculate the maximum value Max of the matching degree Ncc of the overlapping area between the image to be stitched and the stitched image after stitching the two images. Ncc With the set matching degree threshold T N in size. If Max Ncc > T N , mark the image to be stitched as correct; otherwise, mark the image to be stitched as incorrect.

[0071] Among them, the normalized cross-correlation algorithm (Normalized Cross-Correlation, Ncc) accelerated by GPU is used to judge the matching degree Ncc of two images in the overlapping area. The calculation formula of Ncc is:

[0072]

[0073] Among them, f represents the pixel value of the pre-matched image, t represents the pixel value of the template image, μ represents the pixel mean, σ represents the standard deviation, and n represents the total number of pixels.

[0074] 2-2. The method for stitching the first row image, the first column image, and the last column image is as follows: When stitching the first row image, stitch the current image with the previous image in the moving direction. When stitching the first column or the last column image, stitch the current image with the image in the same column as the current image and in the previous row of the current image;

[0075] Among them, if the stitching flag bit of the stitched image is incorrect, directly mark the current image as incorrect; among them, it should be understood that the first row image, the first column image, and the last column image are generally relatively stable and not prone to stitching errors;

[0076] If the stitching flag bit of the stitched image is correct, calculate the maximum value Max of the matching degree Ncc of the overlapping area between the two images after stitching Ncc , if Max Ncc > T N , mark the current image as correct; otherwise, mark the current image as incorrect.

[0077] 2-2. The method for stitching the images at the remaining positions except for the first row image, the first column image, and the last column image is as follows:

[0078] Refer to Figure 3 , record the current image as P0, the previous image in the moving direction of the current image is recorded as P1, and the image in the same column as the previous image and in the previous row of the current image is recorded as P2;

[0079] If one of the images in P1 and P2 is marked as correct and the other is marked as wrong, P0 selects the image marked as correct for stitching, and then calculates the maximum value Max of the matching degree Ncc of the overlapping area between the two images after stitching. Ncc , if Max Ncc > T N , then mark the current image as correct, otherwise mark the current image as wrong;

[0080] If both P1 and P2 are marked as correct, calculate the maximum value Ncc01 of the matching degree of the overlapping area after stitching P0 and P1, and the maximum value Ncc02 of the matching degree of the overlapping area after stitching P0 and P2. When Ncc01 is greater than Ncc02, P0 selects to stitch with P1, and when Ncc01 > T N , mark the current image as correct, otherwise mark the current image as wrong; when Ncc01 is not greater than Ncc02, P0 selects to stitch with P2, and when Ncc02 > T N , mark the current image as correct, otherwise mark the current image as wrong;

[0081] If both P1 and P2 are marked as wrong, P0 performs stitching according to a preset offset and marks the current image as wrong.

[0082] P0 no longer simply stitches with P1 or P2, but selects the image with better stitching effect from these two images for stitching. Doing so can make the stitching more accurate on the one hand, and on the other hand, it can ensure that when there is a stitching error in the stitching process of P1 or P2 itself, P0 can select the image with the correct stitching mark for stitching, thereby avoiding the continuous transmission of errors and limiting them to the current stitching error area.

[0083] 2-3. After determining the image to be stitched, calculate the best displacement vector for stitching the current image with the image to be stitched, and then move the current image to cover the image to be stitched according to the best displacement vector. For example, referring to the figure, assuming the image to be stitched is P1, if the coordinates of the upper left point of the image to be stitched P1 on the stitched whole image are (x1, y1), and the best displacement vector is (x, y), then the coordinates of the upper left point of the image P0 to be stitched on the stitched image are (x1 + x, y1 + y).

[0084] 2-4. For the stitched image, use the gray-scale weighted average fusion algorithm to perform weighted fusion on the overlapping area of the two stitched images, and use the pixel value result after weighted fusion as the pixel value of the overlapping area. The formula of the weighted average fusion algorithm is as follows:

[0085] V = a i ×V 1 +(1 - a i )×V2 , a i = w - i / w;

[0086] Wherein, V is the pixel value after fusion, V 1 , V 2 respectively represent the pixel values of the stitched image and the image to be stitched, a i represents the weight value, w represents the width of fusion, and i represents the abscissa of the current pixel point.

[0087] Referring to FIGS. 4(a)-(f), in an embodiment, it is the stitching process of a 4*4 size skin CT image. The width and height of each image are 1000*1000 pixels. Fig. (a) is the start of stitching, generating a background layer and placing the first image at a fixed position; Fig. (b) is the stitching of the first row of images, and the first row of images is only stitched by matching with the previous image in the current movement direction; Figs. (c) and (e) are the image stitching when changing rows. When changing rows, the current image is only stitched with the corresponding position in the previous row; Fig. (d) is the general stitching process, and the current image is respectively matched with the corresponding position in the previous row and the previous image in the current movement direction, and the image with a higher matching degree is preferably selected for stitching; Fig. (f) is the final result after completing the entire stitching process.

[0088] Embodiment 2

[0089] This embodiment provides a skin CT large field-of-view image stitching system with high reliability and fault tolerance, which uses the method of Embodiment 1 to stitch the skin CT large field-of-view image.

[0090] This embodiment also provides a storage medium, on which a computer program is stored, and when the program is executed, it is used to implement the method of Embodiment 1.

[0091] This embodiment also provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it implements the method of Embodiment 1.

[0092] Although the embodiments of the present invention have been disclosed as above, it is not limited to only the applications listed in the specification and embodiments. It can be fully applied to various fields suitable for the present invention. For those familiar with the field, other modifications can be easily made. Therefore, without departing from the general concept defined by the claims and the equivalent scope, the present invention is not limited to specific details.

Claims

1. A method for stitching large field-of-view skin CT images with high reliability and fault tolerance, characterized in that, it includes: 1) Manually determine the starting point and size of the stitching area, and set M rows × N columns of skin CT images to be stitched in a grid shape, ensuring that there is an overlapping area between two adjacent images on the left and right, and above and below; 2) Make the objective lens of the dermoscope start from the starting point of the stitching area and scan each grid in the first row from left to right in sequence, then switch to the next row and move in the reverse direction, and so on, and complete the scanning of all grids along the S-shaped trajectory; During the scanning process of the objective lens of the dermoscope, first obtain the skin CT image of the current grid, and then stitch the obtained skin CT image onto the large field-of-view skin CT image that has been stitched before scanning the current grid, so as to stitch while obtaining images until the acquisition and stitching of M rows × N columns of skin CT images are completed, and a complete large field-of-view skin CT image is obtained; Among them, assign a stitching flag bit to each image. If the stitching of the image is successful, it is marked as correct, and if the stitching fails, it is marked as incorrect; The method for judging whether the image stitching is successful or not is: When stitching with an image with an incorrect stitching flag bit, the image to be stitched that is stitched with the image to be stitched is directly marked as incorrect; When stitching an image with a correct stitching marker bit, calculate the maximum value Max of the matching degree Ncc of the overlapping area between the image to be stitched and the stitched image after stitching the two images Ncc with the set matching degree threshold T N in size. If Max Ncc > T N , mark the image to be stitched as correct, otherwise mark the image to be stitched as incorrect; The method for stitching the images at the remaining positions except for the first row of images, the first column of images, and the last column of images is: Denote the current image as P0, the previous image in the moving direction of the current image as P1, and the image in the same column as the previous image and in the row above the current image as P2; If one of the images in P1 and P2 is marked as correct and the other is marked as incorrect, P0 selects the image marked as correct for splicing, and then calculates the maximum value Max of the matching degree Ncc of the overlapping area of the two spliced images. Ncc , if Max Ncc > T N , then mark the current image as correct, otherwise mark the current image as incorrect; If both P1 and P2 are marked as correct, calculate the maximum matching degree Ncc01 of the overlapping area after splicing P0 and P1, and the maximum matching degree Ncc02 of the overlapping area after splicing P0 and P2. When Ncc01 is greater than Ncc02, P0 selects to splice with P1, and when Ncc01 > T N , mark the current image as correct, otherwise mark the current image as incorrect; when Ncc01 is not greater than Ncc02, P0 selects to splice with P2, and when Ncc02 > T N , mark the current image as correct, otherwise mark the current image as incorrect; If both P1 and P2 are marked as incorrect, then P0 is stitched according to a preset offset, and the current image is marked as incorrect.

2. The method for stitching large field-of-view skin CT images with high reliability and fault tolerance according to claim 1, characterized in that, In step 2), after obtaining the skin CT image of the current grid, first detect whether the brightness of the current image is within the normal range, and thus judge whether it is necessary to adjust the incident laser intensity of the dermoscope to ensure that an image with a satisfactory brightness is obtained; Among them, the image gray mean value mean and the average deviation mdev are used to judge whether the brightness of the image is within the normal range, and the formula is: mean=∑(x i -ref) / N; mdev = ∑|(i - ref) - mean|×His[i] / ∑His[i]; Among them, the average gray value mean of the image is the average value after subtracting the reference value from the image, where x i is the gray value of the current image pixel. In i - ref, i represents the abscissa value in the gray histogram, and its value range is 0 - 255, ref is the set reference value; N is the number of image pixels; His[i] represents the gray histogram of the image; Set the deviation threshold T according to the characteristics of the skin CT image m , if mdev is greater than T m , it indicates that the brightness is normal; if mdev is less than T m , it indicates that the brightness is abnormal, and judge whether the image is too bright or too dark according to the mean value: if the mean is greater than 0, it means the image is too bright and the laser intensity needs to be reduced; if the mean is less than 0, it means the image is too dark and the laser intensity needs to be increased; after adjusting the laser, re - shoot the skin CT image of the current grid and calculate and judge whether the image brightness is normal until an image with satisfactory brightness is obtained, and then start stitching.

3. The method for stitching large field-of-view skin CT images with high reliability and fault tolerance according to claim 2, characterized in that, The calculation formula for the matching degree Ncc of two images in the overlapping area is: Among them, f represents the pixel value of the pre-matched image, t represents the pixel value of the template image, μ represents the pixel mean value, σ represents the standard deviation, and n represents the total number of pixels.

4. The method for stitching large field-of-view skin CT images with high reliability and fault tolerance according to claim 3, characterized in that, The method for stitching the first row of images, the first column of images, and the last column of images is: when stitching the first row of images, stitch the current image with the previous image in the moving direction; when stitching the first column or the last column of images, stitch the current image with the image in the same column as the current image and in the row above the current image. Among them, if the splicing mark bit of the image to be spliced is incorrect, the current image is directly marked as incorrect; If the splicing marker bit of the spliced image is correct, calculate the maximum value Max of the matching degree Ncc of the overlapping area of the two images after splicing Ncc , if Max Ncc > T N , mark the current image as correct, otherwise mark the current image as incorrect.

5. The method for stitching large field-of-view images of a skin CT with high reliability and fault tolerance according to claim 4, characterized in that, when stitching, calculate the optimal displacement vector of the current image, and then move the current image according to the optimal displacement vector for stitching; For the stitched image, the overlapping area of the two stitched images is weighted and fused through a gray-scale weighted average fusion algorithm, and the pixel value result after weighted fusion is used as the pixel value of the overlapping area. The formula of the weighted average fusion algorithm is as follows: V = a i × V 1 +(1 - a i )× V 2 , a i = w - i / w; Among them, V is the pixel value after fusion, V 1 , V 2 represent the pixel values of the stitched image and the image to be stitched respectively, a i represents the weight value, w represents the fusion width, and i represents the abscissa of the current pixel point.

6. A storage medium, on which a computer program is stored, characterized in that, when the program is executed, it is used to implement the method described in any one of claims 1-5.

7. A computer device, including a memory, a processor, and a computer program stored on the memory and operable on the processor, characterized in that, when the processor executes the computer program, it implements the method described in any one of claims 1-5.

Images