A spinal endoscope image deblurring method and system

Abstract

The present application provides a spinal endoscopic image deblurring method and system, which performs multi-scale convolution operations on spinal endoscopic images to obtain the difference in image contrast between different scales, divides the spinal endoscopic images into multiple local structure images, and determines the brightness uniformity of each local structure image by the difference in pixel distribution gradient and contrast in each local structure image; determines the texture dependency between adjacent local structure images by the texture difference in spinal tissue structure between adjacent local structure images; contrast enhances the brightness of the local structure images according to the dependency and brightness uniformity to obtain a local enhanced image; and fuses and reconstructs the blurred area through all the local enhanced images to obtain a deblurred spinal endoscopic image. The scheme of the present application can achieve local adaptive balanced enhancement of the brightness of spinal endoscopic images, thereby reducing detail blurring caused by uneven illumination.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and more specifically, to a spinal endoscope image deblurring method and system. Background Art

[0002] Image enhancement is a common method in image processing. Image enhancement refers to the process of improving the visual effect of an image or highlighting specific information in an image through digital processing technology to enhance its usability in specific applications. Image enhancement usually includes brightness adjustment, contrast enhancement, edge enhancement, denoising, etc., aiming to improve image quality and make key information clearer. In spinal endoscopic imaging, due to the complex tissue structure and non-uniform lighting conditions inside the human body, images often have problems such as overexposure, shadows, and uneven brightness distribution, which leads to blurred tissue texture details and low contrast in the lesion area, posing challenges to the accurate detection and diagnosis of lesions.

[0003] Existing spinal endoscopy image processing methods mostly focus on global enhancement or single-scale image adjustment. When dealing with uneven illumination and complex texture features, there are often defects such as insufficient optimization of local details or unnatural image contrast. In addition, traditional blur removal methods do not adequately analyze the dependence of texture and structural features, which destroys the texture consistency between adjacent regions and affects the final image quality. However, local adaptive enhancement and blurred area fusion reconstruction can reduce the detail blurring caused by uneven illumination, thereby improving the clarity of spinal endoscopy images. Therefore, how to achieve local adaptive balanced enhancement of the brightness of spinal endoscopy images to reduce the detail blurring caused by uneven illumination has become a difficult problem faced by the industry. Summary of the invention

[0004] The present application provides a spinal endoscope image deblurring method and system, which can achieve local adaptive balanced enhancement of the brightness of the spinal endoscope image, thereby reducing detail blurring caused by uneven illumination.

[0005] In a first aspect, the present application provides a spinal endoscope image deblurring method, comprising the following steps:

[0006] Monitor the process of endoscope probing the spine and collect endoscopic images of the spine during the probing process;

[0007] Performing a multi-scale convolution operation on the texture structure in the spinal endoscopic image to obtain the difference in image contrast between different scales, segmenting the spinal endoscopic image into a plurality of local structure images based on the structural features of the spine, and determining the brightness uniformity of each local structure image through the distribution gradient of pixels in each local structure image and the difference;

[0008] Determine the texture difference of the spinal tissue structure between every two adjacent local structure images, and determine the texture dependency between all adjacent local structure images through all texture differences;

[0009] Contrast-enhance the brightness of each local structure image according to the dependency relationship and each brightness uniformity to obtain a local enhanced image of each local structure image;

[0010] The blurred area in the spinal endoscopic image is fused and reconstructed through all the local enhanced images to obtain a deblurred spinal endoscopic image.

[0011] Preferably, performing a multi-scale convolution operation on the texture structure in the spinal endoscope image to obtain the difference in image contrast between different scales specifically includes:

[0012] Using a Gaussian pyramid to gradually downsample the texture structure in the spinal endoscopic image to generate endoscopic images of different resolutions;

[0013] Convolving each endoscopic image with convolution kernels of different sizes to extract the contrast features of each endoscopic image;

[0014] The difference in image contrast between different scales was determined based on the contrast characteristics of all endoscopic images.

[0015] Preferably, segmenting the spinal endoscopic image into a plurality of local structure images based on the structural features of the spine specifically includes:

[0016] Build a region segmentation model;

[0017] Performing edge detection and morphological processing on the spinal endoscope image to extract contour features and texture features of the spine;

[0018] Setting the contour features and texture features of the spine as guiding parameters of the regional segmentation model;

[0019] The spinal endoscope image is locally segmented using the regional segmentation model to obtain a plurality of local structure images.

[0020] Preferably, determining the brightness uniformity of each local structure image by the distribution gradient of pixels in each local structure image and the difference amount specifically includes:

[0021] Using a gradient operator, a distribution gradient of pixels in each local structure image is determined;

[0022] Performing scale convolution difference compensation on each distribution gradient according to the difference amount to obtain a compensation value of each distribution gradient;

[0023] The brightness uniformity of each local structure image is determined according to the compensation value of each distribution gradient.

[0024] Preferably, determining the texture difference in the spinal tissue structure between every two adjacent local structure images specifically includes:

[0025] For every two adjacent local structure images, the gray level co-occurrence matrix is ​​used to extract the texture information of the two adjacent local structure images on the spinal tissue structure;

[0026] Determine the texture similarity between adjacent local structure images through the texture information;

[0027] The texture difference in the spinal tissue structure between adjacent local structure images is determined according to the texture similarity, and then the texture difference in the spinal tissue structure between every two adjacent local structure images is obtained.

[0028] Preferably, determining the texture dependency between all adjacent local structure images through all texture differences specifically includes:

[0029] Take each local structure image as a graph node;

[0030] Determine the connection relationship between graph nodes based on the texture differences between adjacent local structure images;

[0031] Build a texture dependency graph based on all graph nodes and the connection relationships between all graph nodes;

[0032] The texture dependency graph is used to describe the texture dependency relationship between all adjacent local structure images.

[0033] Preferably, the endoscope is a discoscope.

[0034] In a second aspect, the present application provides a spinal endoscope image deblurring system, comprising:

[0035] An acquisition module, used for monitoring the detection process of the spine by the endoscope and acquiring the endoscopic image of the spine during the detection process;

[0036] A processing module, configured to perform a multi-scale convolution operation on the texture structure in the spinal endoscopic image, thereby obtaining a difference in image contrast between different scales, segmenting the spinal endoscopic image into a plurality of local structure images based on the structural features of the spine, and then determining the brightness uniformity of each local structure image through a distribution gradient of pixels in each local structure image and the difference;

[0037] The processing module is further used to determine the texture difference between every two adjacent local structure images in the spinal tissue structure, and determine the texture dependency between all adjacent local structure images through all texture differences;

[0038] The processing module is further used to perform contrast enhancement on the brightness of each local structure image according to the dependency relationship and each brightness uniformity to obtain a local enhanced image of each local structure image;

[0039] The reconstruction module is used to fuse and reconstruct the blurred area in the spinal endoscopic image through all the local enhanced images to obtain a deblurred spinal endoscopic image.

[0040] In a third aspect, the present application provides a computer device, comprising a memory and a processor, wherein the memory stores codes, and the processor is configured to obtain the codes and execute the above-mentioned spinal endoscopic image deblurring method.

[0041] In a fourth aspect, the present application provides a computer-readable storage medium storing a computer program, which implements the above-mentioned spinal endoscopic image deblurring method when executed by a processor.

[0042] The technical solution provided by the embodiments disclosed in this application has the following beneficial effects:

[0043] In an embodiment of the present application, the detection process of the spine by an endoscope is monitored, and endoscopic images of the spine during the detection process are collected; a multi-scale convolution operation is performed on the texture structure in the endoscopic image of the spine to obtain the difference in image contrast between different scales, and the endoscopic image of the spine is segmented into a plurality of local structure images based on the structural characteristics of the spine, and the brightness uniformity of each local structure image is determined by the distribution gradient of pixels in each local structure image and the difference; the texture difference in the spinal tissue structure between every two adjacent local structure images is determined, and the texture dependency between all adjacent local structure images is determined through all texture differences; the brightness of each local structure image is contrast enhanced according to the dependency and each brightness uniformity to obtain a local enhanced image of each local structure image; the blurred area in the endoscopic image of the spine is fused and reconstructed through all the local enhanced images to obtain a deblurred endoscopic image of the spine.

[0044] It can be seen that the present application performs contrast enhancement on the brightness of each local structure image through dependency and brightness uniformity to obtain a local enhanced image, and then fuses and reconstructs the blurred area in the spinal endoscopic image through all the local enhanced images to obtain a deblurred spinal endoscopic image; firstly, a multi-scale convolution operation is used to extract the contrast difference between different scales, which helps to capture the features of the spinal endoscopic image at different resolutions and enhance the expression ability of complex textures and structures. The brightness uniformity of each local area is further evaluated by combining the distribution gradient and contrast difference of pixels in the local structure image, which can effectively locate the area of ​​uneven illumination; secondly, the texture difference between adjacent local structure images in the spinal tissue structure is used to construct the texture between adjacent local structure images. , according to which the damage to the local texture structure in the traditional blur removal method can be reduced; then, the texture dependency between adjacent local structure images and the local structure image are used to perform contrast enhancement processing on the local structure image, so that the enhancement effect not only maintains the naturalness of the details inside the region, but also can be coordinated with the adjacent regions, thereby achieving a global and local optimization balance; finally, through a processing method that combines global optimization with local compensation, the fuzzy area is fused and reconstructed using a fuzziness weighted mechanism, which can accurately fuse the detail information of the local enhanced image, thereby significantly improving the clarity of the fuzzy area; in summary, the present application scheme can achieve local adaptive balanced enhancement of the brightness of spinal endoscope images, thereby reducing the blurring of details caused by uneven illumination. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 is an exemplary flow chart of a spinal endoscope image deblurring method according to some embodiments of the present application;

[0046] Figure 2 is a schematic diagram of the structure of an endoscope system according to some embodiments of the present application;

[0047] Figure 3 is a schematic diagram of a process for determining texture differences according to some embodiments of the present application;

[0048] Figure 4 is a schematic diagram of the structure of a spinal endoscope image deblurring system according to some embodiments of the present application;

[0049] Figure 5 It is a structural schematic diagram of a computer device for implementing a spinal endoscope image deblurring method according to some embodiments of the present application. DETAILED DESCRIPTION

[0050] In order to better understand the technical solution of the present application, the technical solution of the present application will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0051] refer to Figure 1 , which is an exemplary flow chart of a spinal endoscope image deblurring method according to some embodiments of the present application. The spinal endoscope image deblurring method 100 mainly includes the following steps:

[0052] In step 101, the process of detecting the spine by an endoscope is monitored, and an endoscopic image of the spine during the detection process is acquired.

[0053] It should be noted that an endoscope is a professional imaging instrument used to assist in spinal medical diagnosis. The endoscope is an intervertebral disc endoscope, which enters the patient's spine through a minimally invasive incision to obtain morphological images of the spine and its surrounding tissues. Professional medical staff can use these high-precision images to accurately diagnose the patient's spinal lesions and related conditions. However, due to the limitations of the internal tissue structure of the human body, it is difficult to maintain consistent light intensity and uniformity in different parts of the spine under an endoscope, which can easily lead to overexposure or shadows in the image. The above problems will reduce the display effect of subtle tissues and mild lesions, increase the risk of misdiagnosis or missed diagnosis by inspectors, and thus cause greater difficulties in the diagnosis and treatment of patients.

[0054] It should also be noted that the reference Figure 2 As shown, the figure is a schematic diagram of the structure of the endoscope system in some embodiments of the present application, wherein the endoscope 1 includes an irradiation unit 11, a camera unit 11a and an operating unit 13, which are connected to the control system 2 through the endoscope. The control system includes a control unit 25, a distance estimation unit 26, a determination unit 27 and an image processing unit 21. These parts work together to process the image data from the endoscope. The image processing unit is further connected to the three-dimensional image generation unit 23 and the image synthesis output unit 24, and finally the processed image is displayed on the display device 4. In addition, the light source 3 provides illumination for the endoscope through the light source aperture 31, and the three-dimensional position estimation unit 22 provides spatial positioning information for the system to ensure the accuracy and real-time performance of the image. The entire system realizes the acquisition, processing and three-dimensional visualization of endoscopic images through the orderly collaboration of these modules.

[0055] In the specific implementation, first, during the endoscopic detection process, the posture tracking technology based on inertial sensors (such as Inertial Measurement Unit, IMU) is used to monitor the insertion angle, depth and motion trajectory of the endoscope in real time to ensure the precise positioning of the detection process; then, the real-time video stream acquisition technology is used to use a high-resolution camera to record the internal images of the spine generated during the endoscopic detection process, and each frame of the internal image of the spine is used as a spinal endoscopic image; it should be noted that the present application scheme can adaptively identify the blurriness of each frame of the spinal endoscopic image, and deblur the spinal endoscopic images with higher blurriness.

[0056] In step 102, a multi-scale convolution operation is performed on the texture structure in the spinal endoscopic image to obtain the difference in image contrast between different scales. The spinal endoscopic image is segmented into multiple local structure images based on the structural characteristics of the spine, and the brightness uniformity of each local structure image is determined by the distribution gradient of pixels in each local structure image and the difference.

[0057] In some embodiments, performing a multi-scale convolution operation on the texture structure in the spinal endoscopy image to obtain the difference in image contrast between different scales can be achieved by using the following steps:

[0058] Using a Gaussian pyramid to gradually downsample the texture structure in the spinal endoscopic image to generate endoscopic images of different resolutions;

[0059] Convolving each endoscopic image with convolution kernels of different sizes to extract the contrast features of each endoscopic image;

[0060] The difference in image contrast between different scales was determined based on the contrast characteristics of all endoscopic images.

[0061] In the specific implementation, first, the original spinal endoscope image can be gradually downsampled using a Gaussian pyramid. Each time the spinal endoscope image is blurred by a Gaussian kernel and then downsampled to generate multiple endoscopic images with different resolutions to capture multi-scale information of texture features. For example, assuming that the resolution of the original spinal endoscope image is 1920×1080, the following sub-images with different resolutions can be generated through Gaussian pyramid downsampling: the first layer of downsampling (960×540) reduces the resolution to 1 / 4 of the original (the width and height are halved), retains local texture features, and reduces the high-frequency noise interference of the image; the second layer of downsampling (960×540) reduces the resolution to 1 / 4 of the original (the width and height are halved), retains local texture features, and reduces the high-frequency noise interference of the image; The first layer is downsampled (480×270), and the resolution is further reduced to 1 / 16 of the original, extracting global features in a larger area and ignoring smaller details; the third layer is downsampled (240×135), and the resolution is only 1 / 64 of the original image, which is used to analyze the overall texture and contrast trend of the image; then, convolution operations are performed on each endoscopic image using convolution kernels of different sizes (e.g., 3×3, 5×5, 7×7) to extract the contrast at each resolution; finally, the contrast features of all endoscopic images are summarized, and the absolute difference of all contrast features is used as the difference in image contrast between different scales.

[0062] It should be noted that the contrast difference in this application is an indicator to measure the numerical difference between the contrast features of the same image area at different resolutions, and is used to reflect the degree of detail and texture change of the image at each scale. Its role is to accurately describe the multi-scale texture characteristics in the image by capturing these differences, and provide a basis for identifying uneven lighting, blurred areas, and enhancing key details.

[0063] It should also be noted that the multi-scale convolution in this application can capture texture and contrast features at different scales by extracting and analyzing the multi-resolution characteristics of images in spinal endoscopy image processing, thereby achieving a comprehensive characterization of complex structures. This method enhances the perception of diverse texture information and improves the overall effect of image detail restoration and visual quality.

[0064] In some embodiments, segmenting the spinal endoscopic image into a plurality of local structure images based on the structural features of the spine may be achieved by the following steps:

[0065] Build a region segmentation model;

[0066] Performing edge detection and morphological processing on the spinal endoscope image to extract contour features and texture features of the spine;

[0067] Setting the contour features and texture features of the spine as guiding parameters of the regional segmentation model;

[0068] The spinal endoscope image is locally segmented using the regional segmentation model to obtain a plurality of local structure images.

[0069] In the specific implementation, first, a regional segmentation model is constructed. A segmentation network based on deep learning (such as DeepLab) can be selected. The regional segmentation model can accurately segment complex structures. Secondly, the spinal endoscopic image is preprocessed. The Canny edge detection algorithm can be used to extract the edge information of the spinal contour in the spinal endoscopic image. The extracted edge information is used as the contour feature of the spine, and the texture feature of the spine is extracted in combination with the local directional gradient. Then, the extracted contour features and texture features are used as guiding parameters of the regional segmentation model. Specifically, the contour features of the spine can help the model to clarify the boundaries of the spinal region and reduce the possibility of mis-segmentation, while the texture features of the spine capture the complex details of the spine, which can enhance the model's understanding of the differences within the spinal region. Then, the regional segmentation model's recognition ability of the spinal region is enhanced by parameterization (such as adding constraints to the loss function), which can effectively guide the segmentation process to a more precise direction and avoid fuzzy or incomplete segmentation due to complex data characteristics. Finally, the spinal endoscopic image is input into the regional segmentation model, and the spinal endoscopic image is segmented using the regional segmentation model, and the multiple images obtained by segmentation are used as local structure images.

[0070] In some embodiments, determining the brightness uniformity of each local structure image by the distribution gradient of pixels in each local structure image and the difference amount can be implemented by the following steps:

[0071] Using a gradient operator, a distribution gradient of pixels in each local structure image is determined;

[0072] Performing scale convolution difference compensation on each distribution gradient according to the difference amount to obtain a compensation value of each distribution gradient;

[0073] The brightness uniformity of each local structure image is determined according to the compensation value of each distribution gradient.

[0074] In specific implementation, for each local structure image, first use a gradient operator (such as Sobel) to calculate the gradient value and gradient direction of each pixel in the local structure image, map the gradient values ​​and gradient directions of all pixels into a pixel distribution gradient map, and then use the average value of all gradient amplitudes in the pixel distribution gradient map as the distribution gradient of the pixel in the local structure image; then, the difference amount can be used as a correction factor for the gradient amplitude to reduce the random error introduced by convolutions of different scales and enhance the description characteristics of local brightness changes, and then use the product of the correction factor and the distribution gradient as the compensation value of the distribution gradient. It should be noted that in this application, the difference amount is used to compensate for the distribution gradient, and the gradient value is adjusted through multi-scale convolution operations so that the gradient information can uniformly represent the overall trend of brightness changes; finally, each gradient amplitude in the pixel distribution gradient map is subtracted from the compensation value to obtain the difference between all gradient amplitudes and the compensation value, and then the information variance of all differences is used to describe the brightness uniformity of the local structure image.

[0075] It should be noted that the brightness uniformity in this application is an indicator to measure the brightness smoothness in the local structure image. By quantifying the brightness uniformity, it is possible to effectively describe whether there is obvious uneven illumination in the image, providing a reliable basis for subsequent image enhancement and optimization.

[0076] In step 103, the texture difference in the spinal tissue structure between every two adjacent local structure images is determined, and the texture dependency between all adjacent local structure images is determined through all the texture differences.

[0077] In some embodiments, reference Figure 3 As shown, this figure is a schematic diagram of the process of determining texture differences in some embodiments of the present application. In this embodiment, determining the texture difference in the spinal tissue structure between every two adjacent local structure images can be achieved by using the following steps:

[0078] In step 1031, for every two adjacent local structure images, the texture information of the two adjacent local structure images on the spinal tissue structure is extracted using a gray level co-occurrence matrix;

[0079] In step 1032, the texture similarity between adjacent local structure images is determined by using the texture information;

[0080] In step 1033, the texture difference in the spinal tissue structure between adjacent local structure images is determined according to the texture similarity, and then the texture difference in the spinal tissue structure between every two adjacent local structure images is obtained.

[0081] In the specific implementation, firstly, the gray level co-occurrence matrix is ​​used to extract texture information for every two adjacent local structure images. The texture information specifically includes: texture directionality, contrast, uniformity and other features. When extracting texture information, the gray level co-occurrence matrix of each of the two adjacent local structure images can be calculated by setting different directions (such as 0°, 45°, 90°, 135°) and distance parameters, and the texture information of the adjacent local structure images is obtained from the two gray level co-occurrence matrices; then, the texture similarity of the two adjacent local structure images is calculated using the texture information, and the similarity of the two images in texture features can be quantified by using metrics such as Euclidean distance or cosine similarity; finally, the inverse of the texture similarity can be used as the texture difference. Each can be obtained by the above method to clarify the significance of the texture structure change, and a complete texture difference description matrix can be obtained by summarizing the texture differences of all adjacent local structure images, which can provide an accurate texture dependency relationship basis for subsequent analysis.

[0082] In some embodiments, determining the texture dependency between all adjacent local structure images through all texture differences may be implemented by the following steps:

[0083] Take each local structure image as a graph node;

[0084] Determine the connection relationship between graph nodes based on the texture differences between adjacent local structure images;

[0085] Build a texture dependency graph based on all graph nodes and the connection relationships between all graph nodes;

[0086] The texture dependency graph is used to describe the texture dependency relationship between all adjacent local structure images.

[0087] In the specific implementation, first, each local structure image is defined as a graph node to represent the corresponding image area; secondly, the texture difference (i.e., texture difference value) between adjacent local structure images is used as the connection weight between graph nodes. If the texture difference value is low, the weight is high, indicating a strong dependency relationship, while if the texture difference value is high, the weight is low, indicating a weak dependency relationship; then, graph theory tools (such as the NetworkX library) can be used to connect all graph nodes and weight relationships to construct a weighted undirected graph (i.e., texture dependency graph); finally, the texture dependency graph is used to analyze the texture dependency relationship between adjacent local structure images. For example, the dependency strength of local areas can be identified through characteristics such as the shortest path and the maximum connected component in the graph, which can provide support for subsequent image fusion.

[0088] It should be noted that the dependency relationship in the present application refers to the degree of correlation between the texture features of adjacent local structure images. The texture dependency relationship between adjacent local structure images can ensure the texture consistency of adjacent areas during image enhancement and reconstruction, avoiding the destruction of the overall texture structure due to isolated processing of local images. At the same time, it provides global guidance for the fusion reconstruction of blurred areas, which can improve the structural integrity and visual effect of the deblurred image.

[0089] In step 104, the brightness of each local structure image is contrast-enhanced according to the dependency relationship and each brightness uniformity to obtain a local enhanced image of each local structure image.

[0090] In some embodiments, contrast enhancement is performed on the brightness of each local structure image according to the dependency relationship and each brightness uniformity, and obtaining a local enhanced image of each local structure image can be achieved by using the following steps:

[0091] For each local structure image, determining a low contrast region in the local structure image based on brightness uniformity of the local structure image;

[0092] adjusting the contrast of the low-contrast area according to the dependency;

[0093] The low contrast area is brightness enhanced by using a local histogram equalization method to obtain a local enhanced image of the local structure image, and then a local enhanced image of each local structure image is obtained.

[0094] In specific implementation, for each local structure image, first, by setting a brightness gradient or contrast threshold, the area with low brightness uniformity can be marked as a low contrast area; then, in combination with texture dependency, by adjusting the contrast difference of adjacent local structure images, the brightness and contrast changes between adjacent areas are smoothed, for example, the adjustment coefficient is calculated by the weight of the dependency, and the overall contrast of the low contrast area is gradually improved; finally, the brightness of the low contrast area is enhanced using the local histogram equalization method. Specifically, the grayscale histogram can be calculated for each low contrast area separately, and the pixel value range can be redistributed to enhance the pixel brightness of the low contrast area, thereby improving the visibility of local details, and finally a contrast-enhanced local enhanced image will be generated. The local enhanced images of all local structure images can be obtained in the above manner; it should be noted that the embodiment of the present application makes full use of the information of brightness uniformity and texture dependency to ensure that the enhancement effect takes into account both overall visual consistency and local detail clarity.

[0095] In step 105, the blurred area in the spinal endoscopic image is fused and reconstructed through all the local enhanced images to obtain a deblurred spinal endoscopic image.

[0096] In some embodiments, the blurred area in the spinal endoscopic image is fused and reconstructed by all the local enhanced images to obtain a deblurred spinal endoscopic image, which can be achieved by the following steps:

[0097] Acquire a blurred area in the spinal endoscopy image;

[0098] For each blurred area, a local enhanced image of the area corresponding to the blurred area is obtained;

[0099] Based on the fuzziness weighted fusion mechanism, the local enhanced image and the image in the fuzzy area are fused to obtain a fused image of the fuzzy area, and then the fused images of each fuzzy area are obtained;

[0100] The deblurred spinal endoscopy image is determined by all the fused images.

[0101] In specific implementation, first, the blurred area can be located through a blur detection algorithm (such as Laplace transform) and its boundary information can be recorded; then, for each blurred area, the corresponding local enhanced image is matched, and the local enhanced image and the image in the blurred area are synthesized through a blur weighted fusion mechanism. During the image synthesis process, the blur of the pixels in the blurred area can be calculated. For example, the blur degree of the blurred area is reflected by the gradient amplitude. Pixels with low blur give the enhanced image a higher weight in the fusion. The formula is as follows: fused pixel = w enhancement * enhanced pixel + w blur * blurred pixel, where w enhancement and w blur represent weights, and the weights w enhancement and w blur can be dynamically adjusted based on the size of the blur, and w enhancement + w blur = 1; finally, the fusion results of all blurred areas are spliced ​​into the overall image, and the transition at the junction of the areas is processed by an edge smoothing method (such as Gaussian smoothing) to finally obtain a deblurred spinal endoscopic image.

[0102] On the other hand, in some embodiments, the present application provides a spinal endoscope image deblurring system, referring to Figure 4 , which is a schematic diagram of the structure of a spinal endoscope image deblurring system according to some embodiments of the present application, the spinal endoscope image deblurring system 400 includes: an acquisition module 401, a processing module 402 and a reconstruction module 403, which are respectively described as follows:

[0103] The acquisition module 401 in the present application is mainly used to monitor the detection process of the spine by the endoscope and to acquire the endoscopic image of the spine during the detection process;

[0104] Processing module 402, in the present application, is used to perform a multi-scale convolution operation on the texture structure in the spinal endoscopic image, thereby obtaining the difference in image contrast between different scales, and segmenting the spinal endoscopic image into a plurality of local structure images based on the structural features of the spine, and then determining the brightness uniformity of each local structure image through the distribution gradient of pixels in each local structure image and the difference;

[0105] In the present application, the processing module 402 is also used to determine the texture difference between every two adjacent local structure images in the spinal tissue structure, and determine the texture dependency between all adjacent local structure images through all texture differences;

[0106] In the present application, the processing module 402 is further used to perform contrast enhancement on the brightness of each local structure image according to the dependency relationship and each brightness uniformity to obtain a local enhanced image of each local structure image;

[0107] Reconstruction module 403: In the present application, the reconstruction module 403 is mainly used to fuse and reconstruct the blurred area in the spinal endoscopic image through all the local enhanced images to obtain a deblurred spinal endoscopic image.

[0108] In addition, the present application also provides a computer device, which includes a memory and a processor, wherein the memory stores codes, and the processor is configured to obtain the codes and execute the above-mentioned spinal endoscopic image deblurring method.

[0109] In some embodiments, reference Figure 5 , which is a schematic diagram of the structure of a computer device for implementing a spinal endoscope image deblurring method according to some embodiments of the present application. The spinal endoscope image deblurring method in the above embodiment can be Figure 5 The computer device 500 shown in the figure is implemented, and the computer device 500 includes at least one processor 501, a communication bus 502, a memory 503 and at least one communication interface 504.

[0110] The processor 501 may be a general-purpose central processing unit (CPU) or an application specific integrated circuit (ASIC).

[0111] The communication bus 502 may be used to transmit information between the above-mentioned components.

[0112] The memory 503 may be a read only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, or an electrically erasable programmable read only memory (EEPROM), a compact disc read only memory (CD ROM) or other optical disc storage, an optical disc storage (including a compressed optical disc, a laser disc, an optical disc, a digital versatile disc, a Blu-ray disc, etc.), a magnetic disk or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of an instruction or data structure and can be accessed by a computer, but is not limited thereto. The memory 503 may exist independently and be connected to the processor 501 via the communication bus 502. The memory 503 may also be integrated with the processor 501.

[0113] The memory 503 is used to store the program code for executing the solution of the present application, and the execution is controlled by the processor 501. The processor 501 is used to execute the program code stored in the memory 503. The program code may include one or more software modules. The spinal endoscope image deblurring method in the above embodiment can be implemented by the processor 501 and one or more software modules in the program code in the memory 503.

[0114] The communication interface 504 uses any transceiver or other device for communicating with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc.

[0115] In a specific implementation, as an embodiment, a computer device may include multiple processors, each of which may be a single-core (single CPU) processor or a multi-core (multi CPU) processor. The processor here may refer to one or more devices, circuits, and / or processing cores for processing data (such as computer program instructions).

[0116] The above-mentioned computer device may be a general-purpose computer device or a special-purpose computer device. In a specific implementation, the computer device may be a desktop computer, a portable computer, a network server, a personal digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device or an embedded device. The embodiment of the present application does not limit the type of computer device.

[0117] In addition, the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned spinal endoscopic image deblurring method is implemented.

[0118] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0119] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A spinal endoscope image deblurring method, characterized in that: The steps include: Monitor the process of endoscope probing the spine and collect endoscopic images of the spine during the probing process; Performing a multi-scale convolution operation on the texture structure in the spinal endoscopic image to obtain the difference in image contrast between different scales, segmenting the spinal endoscopic image into a plurality of local structure images based on the structural features of the spine, and determining the brightness uniformity of each local structure image through the distribution gradient of pixels in each local structure image and the difference; Determine the texture difference of the spinal tissue structure between every two adjacent local structure images, and determine the texture dependency between all adjacent local structure images through all texture differences; Contrast-enhance the brightness of each local structure image according to the dependency relationship and each brightness uniformity to obtain a local enhanced image of each local structure image; The blurred area in the spinal endoscopic image is fused and reconstructed through all the local enhanced images to obtain a deblurred spinal endoscopic image.

2. The method according to claim 1, characterized in that The multi-scale convolution operation is performed on the texture structure in the spinal endoscope image to obtain the difference in image contrast between different scales, which specifically includes: Using a Gaussian pyramid to gradually downsample the texture structure in the spinal endoscopic image to generate endoscopic images of different resolutions; Convolving each endoscopic image with convolution kernels of different sizes to extract the contrast features of each endoscopic image; The difference in image contrast between different scales was determined based on the contrast characteristics of all endoscopic images.

3. The method according to claim 1, characterized in that Segmenting the spinal endoscopic image into a plurality of local structure images based on the structural features of the spine specifically includes: Build a region segmentation model; Performing edge detection and morphological processing on the spinal endoscope image to extract contour features and texture features of the spine; Setting the contour features and texture features of the spine as guiding parameters of the regional segmentation model; The spinal endoscope image is locally segmented using the regional segmentation model to obtain a plurality of local structure images.

4. The method according to claim 1, characterized in that Determining the brightness uniformity of each local structure image by the distribution gradient of pixels in each local structure image and the difference amount specifically includes: Using a gradient operator, a distribution gradient of pixels in each local structure image is determined; Performing scale convolution difference compensation on each distribution gradient according to the difference amount to obtain a compensation value of each distribution gradient; The brightness uniformity of each local structure image is determined according to the compensation value of each distribution gradient.

5. The method according to claim 1, characterized in that Determining the texture difference in the spinal tissue structure between each two adjacent local structure images specifically includes: For every two adjacent local structure images, the gray level co-occurrence matrix is ​​used to extract the texture information of the two adjacent local structure images on the spinal tissue structure; Determine the texture similarity between adjacent local structure images through the texture information; The texture difference in the spinal tissue structure between adjacent local structure images is determined according to the texture similarity, and then the texture difference in the spinal tissue structure between every two adjacent local structure images is obtained.

6. The method according to claim 1, characterized in that The texture dependencies between all adjacent local structure images are determined through all texture differences, including: Take each local structure image as a graph node; Determine the connection relationship between graph nodes based on the texture differences between adjacent local structure images; Build a texture dependency graph based on all graph nodes and the connection relationships between all graph nodes; The texture dependency graph is used to describe the texture dependency relationship between all adjacent local structure images.

7. The method according to claim 1, characterized in that The endoscope is a discoscope.

8. A spinal endoscope image deblurring system, characterized in that: include: An acquisition module, used for monitoring the detection process of the spine by the endoscope and acquiring the endoscopic image of the spine during the detection process; A processing module, configured to perform a multi-scale convolution operation on the texture structure in the spinal endoscopic image, thereby obtaining a difference in image contrast between different scales, segmenting the spinal endoscopic image into a plurality of local structure images based on the structural features of the spine, and then determining the brightness uniformity of each local structure image through a distribution gradient of pixels in each local structure image and the difference; The processing module is further used to determine the texture difference between every two adjacent local structure images in the spinal tissue structure, and determine the texture dependency between all adjacent local structure images through all texture differences; The processing module is further used to perform contrast enhancement on the brightness of each local structure image according to the dependency relationship and each brightness uniformity to obtain a local enhanced image of each local structure image; The reconstruction module is used to fuse and reconstruct the blurred area in the spinal endoscopic image through all the local enhanced images to obtain a deblurred spinal endoscopic image.

9. A computer device, comprising a memory and a processor, wherein the memory stores a code, characterized in that: The processor is configured to acquire the code and execute the spinal endoscopy image deblurring method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the spinal endoscope image deblurring method according to any one of claims 1 to 7 is implemented.

Images