Splicing seam repairing method and device, electronic equipment and readable medium

Abstract

The invention relates to a splicing seam repairing method and device, electronic equipment and a readable medium, and the method comprises the steps: obtaining an original DR image collected by DR equipment, and determining a first candidate region corresponding to the position of a grid of the DR equipment on the original DR image; performing multi-algorithm joint calculation on the original DR image based on the first candidate region to determine a splicing seam region in the first candidate region; performing local statistics on the splicing seam region and the neighborhood thereof to obtain local gray feature information, and taking the local gray feature information as a dynamic threshold constraint; and under the constraint of a dynamic threshold value, a weighted diffusion equation is constructed based on the main structure direction and the diffusion weight determined by a structure tensor, iterative repair is carried out on the splicing seam area according to the weighted diffusion equation, and the structure tensor is obtained by calculating gradient information of the splicing seam area and a neighborhood of the splicing seam area. The problem of how to ensure the structure continuity and the detail integrity while repairing the splicing seam of the DR image is solved.

Description

Technical Field

[0001] This application relates to the field of image processing technology, and in particular to a method, apparatus, electronic device, and readable medium for repairing seams. Background Technology

[0002] Medical DR (Digital Radiography) imaging, as a direct digital imaging technology, has been widely used in clinical diagnosis due to its advantages such as high resolution and low radiation dose. However, when acquiring large-size images (e.g., a full-frame 1.2-meter-long plate), current manufacturing limitations prevent the fabrication of a single, monolithic grid to match the image. In practice, two grids are typically stitched together for exposure and imaging. However, this stitching method produces noticeable seams in the resulting DR images, manifesting as uneven brightness, structural breaks, or artifacts. These defects not only affect the overall visual quality of the image but may also interfere with the doctor's accurate observation and diagnosis of lesions, thus reducing the clinical value of medical DR images.

[0003] Existing methods for repairing seams in DR images mainly include interpolation repair, feature matching repair, and deep learning repair. However, these methods still have shortcomings in terms of detail preservation, computational efficiency, structural continuity, and adaptability to large and complex images.

[0004] Therefore, how to repair the seams in DR images while ensuring structural continuity and detail integrity in order to improve the accuracy and practicality of clinical diagnosis has become a key technical problem that urgently needs to be solved.

[0005] There is currently no effective solution to the above problems. Summary of the Invention

[0006] This application provides a method, apparatus, electronic device, and readable medium for repairing seams in DR images, in order to solve the aforementioned technical problem of "how to ensure structural continuity and detail integrity while repairing seams in DR images".

[0007] According to one aspect of the embodiments of this application, this application provides a method for repairing seams, comprising: acquiring an original DR image captured by a DR device, and determining a first candidate region on the original DR image corresponding to the position of the filter grid of the DR device; performing multi-algorithm joint calculation on the original DR image based on the first candidate region to determine a seam region in the first candidate region; performing local statistics on the seam region and its neighborhood to obtain local gray-level feature information, and using the local gray-level feature information as a dynamic threshold constraint; under the dynamic threshold constraint, constructing a weighted diffusion equation based on the principal structure direction determined by the structure tensor and the diffusion weight, and iteratively repairing the seam region according to the weighted diffusion equation, wherein the structure tensor is calculated for the gradient information of the seam region and its neighborhood.

[0008] Optionally, the original DR image is subjected to joint calculation using multiple algorithms based on the first candidate region to determine the stitching seam region in the first candidate region, including: using the first algorithm to calculate the first horizontal gradient and the first vertical gradient of each pixel point in the first candidate region on the original DR image; calculating the gradient magnitude map based on the first horizontal gradient and the first vertical gradient; using the second algorithm to perform threshold segmentation on the gradient magnitude map to obtain the second candidate region; using the third algorithm to refine the edge of the second candidate region to obtain the edge map of the stitching seam; and determining the stitching seam region from the first candidate region through the gradient magnitude map and the edge map.

[0009] Optionally, under dynamic threshold constraints, a weighted diffusion equation is constructed based on the principal structure direction and diffusion weights determined by the structure tensor. This includes: calculating the second horizontal gradient and second vertical gradient of each pixel in the seam region and its neighborhood, wherein each pixel satisfies the dynamic threshold constraints; constructing the structure tensor of each pixel based on the second horizontal gradient and second vertical gradient; performing feature decomposition on the structure tensor to obtain the principal structure direction and gradient change rate; estimating the first pixel value of each pixel in the seam region based on the second horizontal gradient and second vertical gradient; calculating the grayscale change of each pixel in the seam region based on each first pixel value; calculating the diffusion weight of each pixel in the neighborhood based on the grayscale change and gradient change rate; and constructing the weighted diffusion equation using the diffusion weights and the first pixel value.

[0010] Optionally, performing eigenvalue decomposition on the structure tensor to obtain the principal structure direction and gradient rate of change includes: performing eigenvalue decomposition on the structure tensor to obtain a first eigenvalue and a second eigenvalue, wherein each of the first eigenvalue and the second eigenvalue corresponds to an eigenvector; comparing the magnitudes of the first eigenvalue and the second eigenvalue, and determining the larger eigenvalue as the target eigenvalue; taking the direction of the eigenvector corresponding to the target eigenvalue as the principal structure direction, and taking the target eigenvalue as the gradient rate of change.

[0011] Optionally, the grayscale change of each pixel in the seam area is calculated based on each first pixel value, including calculating the grayscale change of the current pixel in the following manner: obtaining the pixel value data of each pixel adjacent to the current pixel, wherein the pixel value data includes the upper pixel value, lower pixel value, left pixel value, and right pixel value; and calculating the grayscale change of the current pixel using the upper pixel value, lower pixel value, left pixel value, and right pixel value, using the following formula: , in, Indicates the pixel value. Indicates the next pixel value. Indicates the right pixel value. This represents the left pixel value.

[0012] Optionally, after obtaining the diffusion weight of the pixel, the method further includes: taking the main structure direction as the diffusion direction, and multiplying the diffusion weight of the pixel by the gradient change rate to obtain the diffusion intensity of the pixel; and driving the pixel value of the splicing seam region to diffuse to the neighborhood according to the diffusion direction and diffusion intensity, so as to repair the splicing seam region.

[0013] Optionally, the seam region is iteratively repaired according to the weighted diffusion equation, including: for the pixel to be repaired, substituting the diffusion weight into the weighted diffusion equation to obtain the second pixel value, wherein the pixel to be repaired is the pixel on the seam region; obtaining the repair result by comparing the first pixel value and the second pixel value of the pixel to be repaired; if the repair result meets the convergence condition, the iteration stops; if the repair result does not meet the convergence condition, the iteration continues until the stopping condition is triggered, wherein the stopping condition is that the repair result meets the convergence condition and / or the cumulative number of iterations equals the maximum number of iterations.

[0014] According to another aspect of the embodiments of this application, this application provides a splicing seam repair device, including: an acquisition module, used to acquire an original DR image collected by a DR device, and determine a first candidate region on the original DR image corresponding to the filter grid position of the DR device; a calculation module, used to perform multi-algorithm joint calculation on the original DR image based on the first candidate region, so as to determine the splicing seam region in the first candidate region; a statistics module, used to perform local statistics on the splicing seam region and its neighborhood to obtain local gray-level feature information, and use the local gray-level feature information as a dynamic threshold constraint; and a repair module, used to construct a weighted diffusion equation based on the principal structure direction determined by the structure tensor and the diffusion weight under the dynamic threshold constraint, and to perform iterative repair on the splicing seam region according to the weighted diffusion equation, wherein the structure tensor is calculated for the gradient information of the splicing seam region and its neighborhood.

[0015] According to another aspect of the embodiments of this application, this application provides an electronic device, including a memory, a processor, a communication interface and a communication bus. The memory stores a computer program that can run on the processor. The memory and the processor communicate with each other through the communication bus and the communication interface. When the processor executes the computer program, it implements the steps of the above method.

[0016] According to another aspect of the embodiments of this application, this application also provides a computer-readable medium having processor-executable non-volatile program code that causes the processor to perform the above-described method.

[0017] Compared with related technologies, the technical solutions provided in this application have the following advantages: This application provides a method for repairing seams in DR images, comprising: acquiring an original DR image captured by a DR device, and determining a first candidate region on the original DR image corresponding to the position of the filter grid of the DR device; performing multi-algorithm joint calculation on the original DR image based on the first candidate region to determine the seam region within the first candidate region; performing local statistics on the seam region and its neighborhood to obtain local gray-level feature information, and using the local gray-level feature information as a dynamic threshold constraint; under the dynamic threshold constraint, constructing a weighted diffusion equation based on the principal structure direction determined by the structure tensor and the diffusion weight, and iteratively repairing the seam region according to the weighted diffusion equation, wherein the structure tensor is calculated based on the gradient information of the seam region and its neighborhood. By introducing a dynamic threshold constraint in the seam region, combining the principal structure direction extracted by the structure tensor with the adaptive diffusion weight, constructing a weighted diffusion equation and iteratively updating pixel values, the repair process follows the local gray-level features of the image while maintaining the continuity of edges and textures, thus solving the problem of how to ensure structural continuity and detail integrity while repairing seams in DR images. Attached Figure Description

[0018] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0019] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, those skilled in the art can obtain other drawings based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of the hardware environment for an optional seam repair method provided according to an embodiment of this application; Figure 2 This is a flowchart of an optional seam repair method provided according to an embodiment of this application; Figure 3 This is a flowchart illustrating an optional method for repairing splice seams of a full-width DR filter grid according to an embodiment of this application; Figure 4 This is a schematic diagram of the framework of an optional full-frame grid image restoration system according to an embodiment of this application; Figure 5 This is a block diagram of an optional seam repair device according to an embodiment of this application; Figure 6 This is a schematic diagram of an optional electronic device structure provided in an embodiment of this application. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0022] In the following description, the use of suffixes such as "module," "part," or "unit" to denote elements is solely for the purpose of illustration and has no specific meaning in itself. Therefore, "module" and "part" may be used interchangeably.

[0023] Medical DR (Digital Radiography) imaging, as a direct digital imaging technology, has been widely used in clinical diagnosis due to its advantages such as high resolution and low radiation dose. However, when acquiring large-size images (e.g., a full-frame 1.2-meter-long plate), current manufacturing limitations prevent the fabrication of a single, monolithic grid to match the image. In practice, two grids are typically stitched together for exposure and imaging. However, this stitching method produces noticeable seams in the resulting DR images, manifesting as uneven brightness, structural breaks, or artifacts. These defects not only affect the overall visual quality of the image but may also interfere with the doctor's accurate observation and diagnosis of lesions, thus reducing the clinical value of medical DR images.

[0024] Existing methods for repairing seams in DR images mainly include interpolation repair, feature matching repair, and deep learning repair. However, these methods still have shortcomings in terms of detail preservation, computational efficiency, structural continuity, and adaptability to large and complex images.

[0025] Therefore, how to repair the seams in DR images while ensuring structural continuity and detail integrity in order to improve the accuracy and practicality of clinical diagnosis has become a key technical problem that urgently needs to be solved.

[0026] To address the problems mentioned in the background art, according to one aspect of the embodiments of this application, an embodiment of a method for repairing splice seams is provided.

[0027] Optionally, in the embodiments of this application, the above-described seam repair method can be applied to, for example... Figure 1 The hardware environment shown consists of terminal 101 and server 103. Figure 1 As shown, server 103 is connected to terminal 101 via a network and can be used to provide services (image processing services, algorithm calculations, etc.) to the terminal or clients installed on the terminal. Database 105 can be set up on the server or independently of the server to provide data storage services for server 103. The network mentioned above includes, but is not limited to, wide area network, metropolitan area network or local area network. Terminal 101 includes, but is not limited to, PC, mobile phone, tablet computer, etc.

[0028] A seam repair method in this application embodiment can be executed by server 103, or it can be executed jointly by server 103 and terminal 101, such as... Figure 2 As shown, it includes: Step 201: Obtain the original DR image acquired by the DR device, and determine the first candidate region on the original DR image corresponding to the position of the filter grid of the DR device; Step 202: Perform multi-algorithm joint calculation on the original DR image based on the first candidate region to determine the splicing seam region in the first candidate region; Step 203: Perform local statistics on the splicing seam area and its neighborhood to obtain local grayscale feature information, and use the local grayscale feature information as a dynamic threshold constraint. Step 204: Under dynamic threshold constraints, a weighted diffusion equation is constructed based on the main structure direction and diffusion weight determined by the structural tensor, and the splicing seam region is iteratively repaired according to the weighted diffusion equation. Here, the structural tensor is calculated for the gradient information of the splicing seam region and its neighborhood.

[0029] This application focuses on solving the core challenges of existing medical DR image stitching seam repair technology. Addressing pain points such as image blurring, artifact retention, loss of detail, low repair accuracy, slow processing efficiency, high data and computing resource consumption, and poor algorithm universality, it proposes a multi-operator collaborative bivariate localization and dynamic constraint gradient-structure tensor diffusion repair scheme.

[0030] By employing multi-operator collaborative localization with Sobel and Otsu methods and dynamic maintenance using bivariate tracking, sub-pixel-level accurate identification of stitching seams is achieved, avoiding false detection of normal tissue edges. Combined with dynamic threshold constraints and an adaptive weighted diffusion mechanism of gradient-structure tensors, defects such as brightness discontinuities and structural fractures at stitching seams are eliminated while preserving high-frequency details such as the bone cortex and trabeculae to the greatest extent possible. Through algorithm optimization and hardware acceleration, the processing time for a single 1.2-meter-long DR image is compressed, eliminating reliance on massive amounts of labeled data, reducing computational resource requirements, and enhancing the algorithm's robustness to imaging differences from different devices, stitching seam morphology variations, and complex anatomical structures. Ultimately, this achieves high-precision, high-efficiency, and highly adaptive repair of stitching seams in medical DR images.

[0031] Figure 3 A flowchart of a method for repairing splice seams of a full-width DR filter grid provided in this application is shown in the figure, including: Step 301: Obtain X-ray images (i.e., raw DR images); Step 302: Preprocess the acquired image; Step 303: The stitching seams of the images are located using a multi-operator collaborative, bivariate tracking method; Step 304: Perform image inpainting on the located image using dynamic threshold constraints and gradient-structure tensor adaptive weighted diffusion.

[0032] The imaging system of a DR device uses X-rays to penetrate human tissue. A flat panel detector receives the X-rays after they have passed through the human body, thereby acquiring the original DR image. This image is stored in RAW format and contains the original data information of the full-frame DR grid splicing seam area.

[0033] Before acquiring raw DR images, the flat panel detector of the DR device is calibrated. Through specific calibration procedures and standard test images, problems such as non-uniform pixel response of the detector are corrected to ensure the accuracy of the acquired images.

[0034] After acquiring the original DR image, the image is preprocessed, including: using a Gaussian filtering algorithm to remove noise from the image, with the Gaussian filtering window size set to 3×3; using histogram equalization technology to enhance image contrast, and redistributing image grayscale values ​​to make image details clearer and improve the overall visual effect.

[0035] For the preprocessed original DR image, the first candidate region of the splicing seam is roughly determined based on the position of the fixed filter grid on the mechanical structure. Then, multi-operator collaboration and bivariate tracking are used to perform fine seam localization to determine the position information of the splicing seam in the original DR image and obtain the splicing seam region.

[0036] By introducing dynamic threshold constraints in the seam region, and combining the principal structure direction extracted from the structural tensor with the adaptive diffusion weight, a weighted diffusion equation is constructed and the pixel values ​​are iteratively updated to complete the repair of the seam region.

[0037] A first candidate region corresponding to the position of the filter grid of the DR device is determined on the original DR image. The position information of the fixed filter grid on the mechanical structure of the DR device is obtained and mapped to the image coordinate system. This region is the first candidate region.

[0038] As an optional embodiment, the original DR image is subjected to joint calculation using multiple algorithms based on a first candidate region to determine the stitching seam region within the first candidate region. This includes: using a first algorithm to calculate the first horizontal gradient and the first vertical gradient of each pixel point in the first candidate region of the original DR image; calculating a gradient magnitude map based on the first horizontal gradient and the first vertical gradient; using a second algorithm to perform threshold segmentation on the gradient magnitude map to obtain a second candidate region; using a third algorithm to refine the edges of the second candidate region to obtain an edge map of the stitching seam; and determining the stitching seam region from the first candidate region using the gradient magnitude map and the edge map.

[0039] The first algorithm is used to calculate the first horizontal gradient and the first vertical gradient of each pixel in the first candidate region of the original DR image. This includes: in the original DR image, for a determined first candidate region, performing a convolution operation on each pixel using the first algorithm (such as the Sobel operator) to obtain the gray-level change rate of that pixel in the horizontal direction, which is used as the first horizontal gradient of that pixel; simultaneously, calculating the gray-level change rate of that pixel in the vertical direction, which is used as the first vertical gradient of that pixel, thereby forming the horizontal gradient map of the first candidate region. With vertical gradient plot .

[0040] According to the first level gradient with the first vertical gradient Calculate gradient magnitude map This includes: for each pixel within the first candidate region, calculating the gradient magnitude of that pixel based on its corresponding first horizontal gradient value and first vertical gradient value using the following formula. : .

[0041] The gradient magnitude values ​​of all pixels are combined to form a gradient magnitude map, which is used to reflect the overall intensity characteristics of grayscale changes within the candidate region.

[0042] After obtaining the gradient magnitude map, a second algorithm is used to process it. This second algorithm is a parallel threshold segmentation method based on the inter-class variance maximization algorithm and the information entropy maximization algorithm. By statistically analyzing the pixel grayscale distribution in the gradient magnitude map and calculating the maximum values ​​of the inter-class variance and information entropy, the optimal segmentation threshold is determined. Based on this optimal segmentation threshold, the gradient magnitude map is binarized to obtain the second candidate region for the stitching seam, which includes the stitching seam region.

[0043] After obtaining the second candidate region, the edges of the second candidate region are refined using a third algorithm. The third algorithm is the Canny operator, which includes Gaussian filtering, non-maximum suppression, and double threshold detection steps: Gaussian filtering removes noise, non-maximum suppression suppresses non-edge point responses, and double threshold detection removes isolated noise points and enhances edge continuity, thereby obtaining an edge map indicating the splicing seam boundary.

[0044] The seam region is determined from the first candidate region using gradient magnitude maps and edge maps. Within the first candidate region, pixels are filtered based on gradient magnitude, and pixels with magnitudes higher than a preset threshold (which can be determined by local statistics or a fixed threshold) are marked as potential seam candidates. The intersection of the potential seam candidates and the edge map is performed, retaining pixels that are both in the high magnitude region and on the edge line, while removing isolated points or non-connected pixels to obtain a joint magnitude-edge candidate set. Connectivity component analysis is performed on the joint magnitude-edge candidate set to remove isolated components with lengths or areas smaller than the threshold, and long strip components are refined (skeletonized) to form a preliminary seam centerline. Using the preliminary seam centerline as the initial position, bivariate tracking is performed on the centerline pixels along the gradient direction and the edge direction to dynamically adjust the position and local shape of the line to enhance continuity and accurately align the seam boundary, thus obtaining the seam region.

[0045] This application uses the Otsu algorithm for threshold segmentation.

[0046] By combining multiple algorithms for computation, gradient magnitude and edge information are effectively utilized to achieve precise positioning of the seam, which can maintain the structural continuity and detail integrity of the DR image, while suppressing isolated noise and artifacts, thereby improving the accuracy of seam repair.

[0047] As an optional embodiment, under dynamic threshold constraints, a weighted diffusion equation is constructed based on the principal structure direction determined by the structure tensor and the diffusion weights. This includes: calculating the second horizontal gradient and second vertical gradient of each pixel in the seam region and its neighborhood, where each pixel satisfies the dynamic threshold constraint; constructing the structure tensor of each pixel based on the second horizontal gradient and second vertical gradient; performing feature decomposition on the structure tensor to obtain the principal structure direction and gradient change rate; estimating the first pixel value of each pixel in the seam region based on the second horizontal gradient and second vertical gradient; calculating the grayscale change of each pixel in the seam region based on each first pixel value; calculating the diffusion weight of each pixel in the neighborhood based on the grayscale change and the gradient change rate; and constructing the weighted diffusion equation using the diffusion weights and the first pixel value.

[0048] The seam area is divided into local windows (such as 3×3 or 5×5 pixels). The mean, variance, range and other statistics of the pixel values ​​in each window are calculated. Then, the local grayscale feature information of each window is extracted.

[0049] Using local grayscale feature information as a dynamic threshold constraint includes: dynamically generating a threshold based on local grayscale features to determine whether a pixel needs to participate in the repair and adjusting the intensity of the weighted diffusion.

[0050] Calculate the horizontal gradient of the seam region and its neighboring pixels. and vertical gradient Based on horizontal gradient and vertical gradient Construct the structure tensor using the following formula: .

[0051] Based on the structural tensor, eigenvalue decomposition is performed to obtain the local principal structure direction and gradient rate of change.

[0052] The first pixel value of each pixel in the seam region can be estimated based on the second horizontal gradient and the second vertical gradient. Specifically, this involves: calculating the second horizontal gradient and the second vertical gradient of each pixel in the seam region using the central difference method; normalizing the second horizontal gradient and the second vertical gradient to obtain the direction vector of the pixel; and using this direction vector as the first pixel value to characterize the local structural change trend, providing a basis for subsequent calculations of diffusion direction and diffusion intensity.

[0053] The grayscale change of each pixel in the seam area is calculated based on the first pixel value, including: updating the pixel value of the current pixel by using the first pixel values ​​of the 8 pixels adjacent to the current pixel, and updating it as the grayscale change. Here, the 8 pixels refer to the 8 pixels surrounding the current pixel.

[0054] Based on the amount of grayscale change And the gradient change rate calculation calculates the diffusion weight W of pixels in each direction in the neighborhood. D The calculation formula is: , Where D represents direction. , representing the four directions: up, down, left, and right. This represents the rate of change of the gradient in direction D. This represents the amount of grayscale change in direction D.

[0055] A weighted diffusion equation is constructed using the diffusion weights and the first pixel value. Let the center pixel to be updated be... Its neighboring pixels are pixels in the four directions: up, down, left, and right. The first pixel value and diffusion weight of the pixel in the up direction are denoted as follows: With W n The first pixel value and diffusion weight of the downward-direction pixel are denoted as follows: With W s The first pixel value and diffusion weight of the left-hand pixel are denoted as follows: With W w The first pixel value and diffusion weight of the pixel to the right are denoted as follows: With W e Then the center pixel The weighted diffusion update formula can be expressed as: .

[0056] This embodiment combines dynamic threshold constraints with the principal direction of the structural tensor to guide pixels to perform weighted diffusion along the main local structure of the image. At the same time, it uses neighborhood adaptive weights to control the diffusion intensity, thereby achieving precise iterative updates of pixels in the stitching seam area and effectively eliminating brightness faults and structural breaks.

[0057] As an optional embodiment, eigenvalue decomposition is performed on the structure tensor to obtain the principal structure direction and gradient rate of change, including: performing eigenvalue decomposition on the structure tensor to obtain a first eigenvalue and a second eigenvalue, wherein each of the first eigenvalue and the second eigenvalue corresponds to an eigenvector; comparing the magnitudes of the first eigenvalue and the second eigenvalue, and determining the larger eigenvalue as the target eigenvalue; taking the direction of the eigenvector corresponding to the target eigenvalue as the principal structure direction, and taking the target eigenvalue as the gradient rate of change.

[0058] The structure tensor in this embodiment is: .

[0059] Performing eigenvalue decomposition on the structure tensor yields two eigenvalues: and These correspond to two feature vectors: and Assuming > ,but The vector direction is the main structural direction, indicating that pixels need to spread along this direction. The rate of change of the gradient in the main direction, Let be the gradient rate of change in the secondary direction.

[0060] It should be noted that, if ≈ (If the difference between the two is less than the set value, and the set value is small enough, it means that the texture of the splicing seam area is uniform with the surrounding texture, and the diffusion can proceed in multiple directions.)

[0061] Alternatively, ( - As the principal gradient rate of change, the larger the value, the stronger the directionality of the region (such as a clear edge). If the value is close to 0, it means that the region is flat and has weak texture.

[0062] By performing eigenvalue decomposition on the structure tensor and selecting the largest eigenvalue and its corresponding eigenvector, the dominant direction information of the image can be accurately captured in the local neighborhood. This direction is then used as the dominant diffusion direction, and the gradient change rate is quantized with the corresponding eigenvalue.

[0063] The restoration process ensures the continuity of edges and texture structures while avoiding the blurring caused by isotropic diffusion, thereby effectively improving the restoration accuracy and image detail fidelity of the seam area.

[0064] As an optional embodiment, the grayscale change of each pixel in the seam area is calculated based on each first pixel value, including calculating the grayscale change of the current pixel in the following manner: obtaining pixel value data of each pixel adjacent to the current pixel, wherein the pixel value data includes the upper pixel value, lower pixel value, left pixel value, and right pixel value; and calculating the grayscale change of the current pixel using the upper pixel value, lower pixel value, left pixel value, and right pixel value, using the following formula: , in, Indicates the pixel value. Indicates the next pixel value. Indicates the right pixel value. This represents the left pixel value.

[0065] Unlike traditional linear interpolation methods that rely solely on pixels on both sides of the seam for estimation, this embodiment calculates grayscale changes by combining the pixel values ​​of neighboring pixels above, below, left, and right. This allows for a more comprehensive utilization of local structural information, improving not only the smoothness and stability of the repair results but also better maintaining structural continuity and detail integrity. This effectively avoids the blurring and structural breakage problems common in interpolation methods.

[0066] As an optional embodiment, after obtaining the diffusion weight of the pixel, the method further includes: taking the main structure direction as the diffusion direction, and multiplying the diffusion weight of the pixel by the gradient change rate to obtain the diffusion intensity of the pixel; and driving the pixel value of the splicing seam region to diffuse to the neighborhood according to the diffusion direction and diffusion intensity, so as to repair the splicing seam region.

[0067] In each iteration, the pixel values ​​are adjusted by weighted averaging to gradually approximate the normal structure and texture features of the neighborhood, thus achieving natural and seamless seam repair.

[0068] The principal structure direction obtained through structural tensor feature decomposition accurately reflects the main texture direction of the image in local regions. Combined with the diffusion intensity obtained by multiplying the diffusion weights by the gradient change rate, the directionality and strength of pixel value diffusion can be simultaneously characterized. Using the principal structure direction as the diffusion path and the diffusion intensity as an adjustment factor, the pixel values ​​in the seam region are diffused to the neighborhood along directions with strong structural continuity. This allows the restoration process to follow the structural features of the original image, achieving natural content filling. The restored seam region maintains structural continuity with the surrounding normal areas while avoiding distortion or breakage of the texture direction.

[0069] As an optional embodiment, iterative repair of the seam region is performed according to the weighted diffusion equation, including: for the pixel to be repaired, substituting the diffusion weight into the weighted diffusion equation to obtain the second pixel value, wherein the pixel to be repaired is the pixel on the seam region; obtaining the repair result by comparing the first pixel value and the second pixel value of the pixel to be repaired; if the repair result meets the convergence condition, the iteration stops; if the repair result does not meet the convergence condition, the iteration continues until the stopping condition is triggered, wherein the stopping condition is that the repair result meets the convergence condition and / or the cumulative number of iterations is equal to the maximum number of iterations.

[0070] The convergence condition for the repair result is that the change in pixel values ​​before and after the repair is less than a preset threshold. This application does not limit the specific value of the preset threshold.

[0071] During the iteration process, the change in pixel value is continuously calculated. When the change is less than a preset threshold or the maximum number of iterations is reached, the iteration is terminated and the repair is completed.

[0072] For example, setting the maximum number of iterations. In each iteration, an adaptive weight is used to perform a weighted average of the weighted diffusion equation, updating the pixel values ​​of the repaired region. This process gradually approximates the structure and texture features of the surrounding normal area. After each iteration, the pixel values ​​of the repaired area are calculated and compared with the result of the previous iteration. The iteration stops when the repair result converges. If the repair result does not converge, the next round of repair continues, the second horizontal gradient and the second vertical gradient are recalculated, and then the same operation as described above is performed.

[0073] By substituting diffusion weights into the weighted diffusion equation and iteratively updating the pixel values ​​in the seam region, the structural and texture features of the neighboring normal pixels can be gradually approximated, achieving natural and continuous repair of the seam. Simultaneously, convergence conditions and a maximum number of iterations are used to control the iteration process, ensuring repair accuracy while preventing excessive iteration.

[0074] After completing the iterative restoration, the restored image is filtered and optimized, including median filtering, neighborhood difference, and smooth transition, to obtain a high-quality restored image.

[0075] Median filtering includes performing median filtering on the repaired area to effectively remove random noise that may be introduced during the iterative diffusion process, thus maintaining the smoothness and stability of the image.

[0076] Neighborhood difference calculation involves calculating the difference between a pixel within the repair area and its neighboring pixels, and using the difference results to enhance local edge details, thereby improving edge clarity and structural continuity.

[0077] Smooth transition includes: through the joint optimization of filtering and differential processing, the transition between the repaired area and the surrounding normal area is made more natural and smooth.

[0078] By introducing joint optimization of median filtering and neighborhood difference after iterative restoration, noise introduced during the restoration process can be effectively removed, edge details can be enhanced, and a natural and smooth transition between the restored area and the surrounding area can be achieved, thereby obtaining a high-quality, structurally continuous final restored image.

[0079] This application provides a method for repairing seams in DR images, comprising: acquiring an original DR image captured by a DR device, and determining a first candidate region on the original DR image corresponding to the position of the filter grid of the DR device; performing multi-algorithm joint calculation on the original DR image based on the first candidate region to determine the seam region within the first candidate region; performing local statistics on the seam region and its neighborhood to obtain local gray-level feature information, and using the local gray-level feature information as a dynamic threshold constraint; under the dynamic threshold constraint, constructing a weighted diffusion equation based on the principal structure direction determined by the structure tensor and the diffusion weight, and iteratively repairing the seam region according to the weighted diffusion equation, wherein the structure tensor is calculated based on the gradient information of the seam region and its neighborhood. By introducing a dynamic threshold constraint in the seam region, combining the principal structure direction extracted by the structure tensor with the adaptive diffusion weight, constructing a weighted diffusion equation and iteratively updating pixel values, the repair process follows the local gray-level features of the image while maintaining the continuity of edges and textures, thus solving the problem of how to ensure structural continuity and detail integrity while repairing seams in DR images.

[0080] According to another aspect of the embodiments of this application, this application also provides an architecture diagram of a full-frame raster image restoration system, such as... Figure 4 As shown, it includes: an image preprocessing module 401, a seam positioning module 402, an image restoration module 403, and a post-processing module 404, wherein the functions of each module include: The image preprocessing module integrates Gaussian filtering and histogram equalization algorithms. After receiving the original medical DR image, it sequentially performs noise removal and contrast enhancement operations. Gaussian filtering suppresses impulse noise in the image, and histogram equalization enhances the overall image contrast, outputting a high-quality preprocessed image that lays the foundation for subsequent processing.

[0081] The seam positioning module includes sub-functions such as flat panel calibration, image acquisition, fixed grid position acquisition, multi-operator collaboration, and bivariate tracking positioning. Flat panel calibration ensures the accuracy of image acquisition. After acquiring image and grid position information, multiple operators are used collaboratively to locate the seam. The bivariate tracking method outputs the precise position information of the seam, providing accurate location guidance for subsequent repair.

[0082] The image restoration module, based on the location information output by the multi-operator, bivariate method gap location module, performs operations such as depth feature extraction, adaptive weighted diffusion, and iterative convergence restoration on the splicing seam area to achieve high-precision restoration of the splicing seam and restore the integrity and naturalness of the image.

[0083] The post-processing module receives the restored image and performs filtering optimization using a combination of median filtering and neighborhood differencing. Median filtering eliminates residual noise from the restoration process, while neighborhood differencing enhances image edges, optimizes the transition between the restored area and the surrounding image, and outputs a final high-quality restored image.

[0084] The data interaction mechanism in the system architecture is responsible for data transmission and interaction between various functional modules, ensuring the accurate and efficient transfer of raw image data, intermediate calculation results, and location information between different modules. Simultaneously, it enables data communication with DR equipment, receiving raw image data and outputting restoration results to designated storage locations or display devices, achieving seamless integration with medical image archiving and communication systems.

[0085] At the hardware level, the system can be integrated into the data processing unit of the DR device, using a GPU (Graphics Processing Unit) to accelerate the calculation of gradients, structure tensors, and cross-diffusion processes. Leveraging the parallel computing capabilities of the GPU significantly improves computational efficiency, meeting the needs of real-time clinical diagnosis. At the software level, the software acquisition module supports RAW format image input, and the restoration results automatically embed original image metadata, facilitating image storage, management, and retrieval.

[0086] According to another aspect of the embodiments of this application, this application provides a splice seam repair device, such as... Figure 5 As shown, it includes: The acquisition module 501 is used to acquire the original DR image captured by the DR device and determine the first candidate region on the original DR image corresponding to the position of the filter grid of the DR device; The calculation module 502 is used to perform multi-algorithm joint calculation on the original DR image based on the first candidate region in order to determine the stitching seam region in the first candidate region; The statistics module 503 is used to perform local statistics on the splicing seam area and its neighborhood to obtain local gray-scale feature information, and use the local gray-scale feature information as a dynamic threshold constraint. Repair module 504 is used to construct a weighted diffusion equation based on the main structure direction and diffusion weight determined by the structure tensor under dynamic threshold constraints, and to iteratively repair the splicing seam region according to the weighted diffusion equation. The structure tensor is calculated for the gradient information of the splicing seam region and its neighborhood.

[0087] It should be noted that the acquisition module 501 in this embodiment can be used to execute step 201 in this application embodiment, the calculation module 502 in this embodiment can be used to execute step 202 in this application embodiment, the statistics module 503 in this embodiment can be used to execute step 203 in this application embodiment, and the repair module 504 in this embodiment can be used to execute step 204 in this application embodiment.

[0088] Optionally, the calculation module 502 is specifically used to calculate the first horizontal gradient and the first vertical gradient of each pixel point in the first candidate region on the original DR image using the first algorithm; calculate the gradient magnitude map based on the first horizontal gradient and the first vertical gradient; perform threshold segmentation on the gradient magnitude map using the second algorithm to obtain the second candidate region; refine the edge of the second candidate region using the third algorithm to obtain the edge map of the stitching seam; and determine the stitching seam region from the first candidate region through the gradient magnitude map and the edge map.

[0089] Optionally, the repair module 504 includes: The first unit is used to calculate the second horizontal gradient and the second vertical gradient of each pixel in the splicing seam region and its neighborhood, wherein each pixel satisfies the dynamic threshold constraint. The second unit is used to construct the structure tensor of each pixel based on the second horizontal gradient and the second vertical gradient. The third unit is used to perform eigenvalue decomposition on the structure tensor to obtain the principal structure direction and gradient rate of change. The fourth unit is used to estimate the first pixel value of each pixel point in the seam area based on the second horizontal gradient and the second vertical gradient. The fifth unit is used to calculate the grayscale change of each pixel in the seam area based on the first pixel value. The sixth unit is used to calculate the diffusion weight of each pixel in the neighborhood based on the grayscale change and the gradient change rate. Unit 7 is used to construct a weighted diffusion equation using diffusion weights and grayscale changes.

[0090] Optionally, the third unit is also used to perform eigenvalue decomposition on the structure tensor to obtain a first eigenvalue and a second eigenvalue, wherein each of the first eigenvalue and the second eigenvalue corresponds to an eigenvector; compare the magnitudes of the first eigenvalue and the second eigenvalue, and determine the larger eigenvalue as the target eigenvalue; take the direction of the eigenvector corresponding to the target eigenvalue as the principal structure direction, and take the target eigenvalue as the gradient rate of change.

[0091] Optionally, the fifth unit is also used to calculate the grayscale change of the current pixel in the following manner: obtaining the pixel value data of each pixel adjacent to the current pixel, wherein the pixel value data includes the upper pixel value, lower pixel value, left pixel value, and right pixel value; and calculating the grayscale change of the current pixel using the upper pixel value, lower pixel value, left pixel value, and right pixel value, using the following formula: , in, Indicates the pixel value. Indicates the next pixel value. Indicates the right pixel value. This represents the left pixel value.

[0092] Optionally, the repair module 504 is further configured to, after obtaining the diffusion weight of the pixel, take the main structure direction as the diffusion direction, and multiply the diffusion weight of the pixel by the gradient change rate to obtain the diffusion intensity of the pixel; and drive the pixel value of the splicing seam region to diffuse to the neighborhood according to the diffusion direction and diffusion intensity, so as to repair the splicing seam region.

[0093] Optionally, the repair module 504 is further configured to substitute the diffusion weight into the weighted diffusion equation for the pixel to be repaired to obtain a second pixel value, wherein the pixel to be repaired is a pixel on the seam area; by comparing the first pixel value and the second pixel value of the pixel to be repaired, a repair result is obtained; if the repair result satisfies the convergence condition, the iteration stops; if the repair result does not satisfy the convergence condition, the iteration continues until a stopping condition is triggered, wherein the stopping condition is that the repair result satisfies the convergence condition and / or the cumulative number of iterations is equal to the maximum number of iterations.

[0094] It should be noted that the examples and application scenarios implemented by the above modules and corresponding steps are the same, but are not limited to the content disclosed in the above embodiments. It should also be noted that the above modules, as part of a device, can operate in environments such as... Figure 1 The hardware environment shown.

[0095] According to another aspect of the embodiments of this application, this application provides an electronic device for implementing the above-described seam repair method, such as... Figure 6 As shown, the electronic device 600 includes a processor 601, a communication bus 602, a user interface 603, and a memory 604.

[0096] As the core computing component of electronic devices, the processor employs a combination of a high-performance central processing unit (CPU) and a graphics processing unit (GPU). The CPU is responsible for executing system control programs and logical judgments, while the GPU utilizes its powerful parallel computing capabilities to accelerate computationally intensive tasks such as gradient calculation, structural tensor decomposition, and weighted diffusion repair, ensuring the efficient operation of the repair system.

[0097] The communication bus is used to connect components such as processors, memory, and user interfaces to achieve high-speed data transmission. It supports multiple communication protocols and can stably and quickly transmit image data, control commands, and other information, ensuring the coordinated operation of various components within the electronic device.

[0098] The user interface includes input / output devices such as a display screen, keyboard, and mouse, providing an operating interface for users. Users can view the original image and restoration results on the monitor, and set system parameters such as contrast, sharpness, and image restoration level parameters using the keyboard and mouse, enabling human-computer interaction and facilitating user monitoring and adjustment of the restoration process.

[0099] The memory includes random access memory (RAM) and non-volatile memory (such as hard disks and solid-state drives). RAM is used to temporarily store image data being processed, intermediate calculation results, and running program code, ensuring fast data reading and writing and improving system operating efficiency. Non-volatile memory is used to store system programs, algorithm code, user settings parameters, and historical repaired image data, allowing users to access and view them at any time, while ensuring that data is not lost after the system restarts.

[0100] According to another aspect of the embodiments of this application, a computer-readable medium having processor-executable non-volatile program code is also provided.

[0101] Optionally, specific examples in this embodiment can refer to the examples described in the above embodiments, and will not be repeated here.

[0102] In specific implementation, the embodiments of this application can be referred to the above embodiments and have corresponding technical effects.

[0103] It is understood that the embodiments described herein can be implemented in hardware, software, firmware, middleware, microcode, or a combination thereof. For hardware implementation, the processing unit can be implemented in one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), general-purpose processors, controllers, microcontrollers, microprocessors, other electronic units for performing the functions described herein, or combinations thereof.

[0104] For software implementation, the techniques described herein can be implemented by units that perform the functions described herein. The software code can be stored in memory and executed by a processor. The memory can be implemented in the processor or external to the processor.

[0105] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0106] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0107] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0108] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0109] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0110] If the aforementioned function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiments of this application, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks. It should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. In the absence of further restrictions, an element defined by the phrase "comprising a..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0111] The above description is merely a specific embodiment of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A method for repairing splice seams, characterized in that, include: Acquire the original DR image captured by the DR device, and determine the first candidate region on the original DR image that corresponds to the position of the filter grid of the DR device; Based on the first candidate region, the original DR image is subjected to joint calculation using multiple algorithms to determine the stitching seam region within the first candidate region; Local statistics are performed on the splicing seam area and its neighborhood to obtain local grayscale feature information, and the local grayscale feature information is used as a dynamic threshold constraint. Under the dynamic threshold constraint, a weighted diffusion equation is constructed based on the main structure direction determined by the structural tensor and the diffusion weight, and the splicing seam region is iteratively repaired according to the weighted diffusion equation. The structural tensor is calculated for the gradient information of the splicing seam region and its neighborhood.

2. The method according to claim 1, characterized in that, The step of performing multi-algorithm joint calculations on the original DR image based on the first candidate region to determine the seam region within the first candidate region includes: The first algorithm is used to calculate the first horizontal gradient and the first vertical gradient of each pixel in the first candidate region of the original DR image. Calculate the gradient magnitude map based on the first horizontal gradient and the first vertical gradient; The gradient magnitude map is thresholded using a second algorithm to obtain a second candidate region. The edge of the second candidate region is refined using the third algorithm to obtain the edge map of the splicing seam; The seam region is determined from the first candidate region using the gradient magnitude map and the edge map.

3. The method according to claim 1, characterized in that, The construction of a weighted diffusion equation based on the principal structure direction determined by the structure tensor and the diffusion weight under the dynamic threshold constraint includes: Calculate the second horizontal gradient and the second vertical gradient of each pixel in the splicing seam region and the neighborhood, wherein each pixel satisfies the dynamic threshold constraint; The structural tensor of each pixel is constructed based on the second horizontal gradient and the second vertical gradient; Eigenvalue decomposition is performed on the structural tensor to obtain the principal structure direction and gradient rate of change. The first pixel value of each pixel point on the seam region is estimated based on the second horizontal gradient and the second vertical gradient; Calculate the grayscale change of each pixel in the splicing seam area based on each of the first pixel values; The diffusion weight of each pixel in the neighborhood is calculated based on the grayscale change and the gradient change rate. The weighted diffusion equation is constructed using the diffusion weights and the first pixel value.

4. The method according to claim 3, characterized in that, The step of performing eigenvalue decomposition on the structural tensor to obtain the principal structure direction and gradient rate of change includes: The structural tensor is subjected to eigenvalue decomposition to obtain a first eigenvalue and a second eigenvalue, wherein each of the first eigenvalue and the second eigenvalue corresponds to an eigenvector; Compare the first feature value with the second feature value, and determine the larger feature value as the target feature value; The direction of the feature vector corresponding to the target feature value is taken as the main structure direction, and the target feature value is taken as the gradient rate of change.

5. The method according to claim 3, characterized in that, The step of calculating the grayscale change of each pixel in the seam region based on each of the first pixel values ​​includes calculating the grayscale change of the current pixel in the following manner: Obtain pixel value data of each pixel adjacent to the current pixel, wherein the pixel value data includes the upper pixel value, lower pixel value, left pixel value, and right pixel value; The grayscale change of the current pixel is calculated using the upper pixel value, the lower pixel value, the left pixel value, and the right pixel value. The calculation formula is as follows: ， in, This indicates the pixel value. This indicates the lower pixel value. This represents the right pixel value. This represents the left pixel value.

6. The method according to claim 3, characterized in that, After obtaining the diffusion weight of the pixel, the method further includes: The diffusion intensity of the pixel is obtained by taking the direction of the main structure as the diffusion direction and multiplying the diffusion weight of the pixel by the gradient change rate. Based on the diffusion direction and the diffusion intensity, the pixel values ​​of the seam region are driven to diffuse to the neighborhood to repair the seam region.

7. The method according to claim 3, characterized in that, The iterative repair of the seam area based on the weighted diffusion equation includes: For the pixel to be repaired, the diffusion weight is substituted into the weighted diffusion equation to obtain the second pixel value, wherein the pixel to be repaired is the pixel on the splicing seam area; The repair result is obtained by comparing the first pixel value and the second pixel value of the pixel to be repaired; If the repair result satisfies the convergence condition, the iteration stops; if the repair result does not satisfy the convergence condition, the iteration continues until a stopping condition is triggered. The stopping condition is that the repair result satisfies the convergence condition and / or the cumulative number of iterations equals the maximum number of iterations.

8. A splice seam repair device, characterized in that, include: The acquisition module is used to acquire the original DR image captured by the DR device, and determine a first candidate region on the original DR image that corresponds to the position of the filter grid of the DR device; The calculation module is used to perform multi-algorithm joint calculation on the original DR image based on the first candidate region, so as to determine the stitching seam region in the first candidate region; The statistics module is used to perform local statistics on the splicing seam area and its neighborhood to obtain local grayscale feature information, and use the local grayscale feature information as a dynamic threshold constraint. The repair module is used to construct a weighted diffusion equation based on the main structure direction and diffusion weight determined by the structure tensor under the dynamic threshold constraint, and to iteratively repair the splicing seam region according to the weighted diffusion equation, wherein the structure tensor is calculated for the gradient information of the splicing seam region and its neighborhood.

9. An electronic device comprising a memory, a processor, a communication interface, and a communication bus, wherein the memory stores a computer program executable on the processor, and the memory and the processor communicate via the communication bus and the communication interface, characterized in that... When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable medium having processor-executable non-volatile program code, characterized in that, The program code causes the processor to execute the method of any one of claims 1 to 7.

Images