Mine image detail enhancement algorithm based on random walk

Abstract

The invention belongs to the field of image detail enhancement, and particularly relates to an underground mine image detail enhancement algorithm based on random walk, which comprises the following steps: obtaining a smooth layer I2 and a detail layer I1 of an original image I0 through two times of bilinear interpolation separation, and calculating an initial residual feature Res; initializing a starting point position of random walk, namely an original image block, and a node position, namely a to-be-matched image block, and calculating a loss function of an optimal matching image block and the original image block; taking minimization of a loss function as a target, searching for a better matching image block, and continuously updating a migration position; and carrying out continuous iteration through random walk, gradually converging to a global optimal matching image block, finishing updating of residual features, amplifying the updated residual, and superposing the updated residual with an input original image to obtain a final detail enhanced image. According to the method, a random walk idea is innovatively introduced, a new residual feature updating mechanism is provided, and the detail enhancement effect is optimized through global search and image block matching.

Description

Technical Field

[0001] The present invention belongs to the field of image detail enhancement, and in particular relates to a random walk-based underground mine image detail enhancement algorithm. Background Art

[0002] When monitoring equipment collects images underground, uneven lighting, high levels of suspended air impurities, and other factors often lead to reduced image contrast and significantly reduced detail recognition, hindering the application of intelligent detection technology. However, these issues have been effectively alleviated by introducing an image detail enhancement algorithm. This algorithm not only optimizes the overall color distribution of the image but also performs refined processing on localized areas, ensuring that key information within the image is clearly presented. This technology has greatly improved the usability of underground mine image data, providing a more reliable monitoring method for safe production.

[0003] Most existing algorithms suffer from issues such as gradient reversal artifacts, increased noise, and color deviation. Some also have poor generalization capabilities and cannot adapt to complex scenarios. Greedy algorithms that search for matching similar image patches to update residual features often tend to only accept the best state, leading to a local optimum. This limits the effectiveness of image detail enhancement. These issues collectively limit the performance of current algorithms in the complex, dark, and highly noisy environments of underground mines, hindering the accuracy of safety monitoring equipment in identifying potential hazards and posing a challenge to safe production in mines. Summary of the Invention

[0004] The technical problem to be solved by the present invention is that the existing algorithm model easily falls into a local optimum, which limits the effect of image detail enhancement.

[0005] To this end, the present invention provides a mine image detail enhancement algorithm based on random walk.

[0006] The technical solution adopted by the present invention to solve its technical problem is:

[0007] A mine image detail enhancement algorithm based on random walk, including:

[0008] Step 1: obtain the smooth layer I2 and detail layer I1 of the original image I0 through two bilinear interpolation separations, and calculate the initial residual feature Res;

[0009] Step 2: Initialize the "starting point" position of the random walk, which is the original image block, and the "node position" is the image block to be matched, and calculate the loss function of the best matching image block and the original image block;

[0010] Step 3: With the goal of minimizing the loss function, find a better matching image block and continuously update the wandering position;

[0011] Step 4: Through continuous iteration of random walks, the algorithm gradually converges to the globally optimal matching image block, completes the update of the residual features, amplifies the updated residual and superimposes it with the input original image to obtain the final detail-enhanced image.

[0012] Furthermore, in the step 1, the residual Res of the original image I0 is defined as I0-I2, and the smoothing layer I2 of the original image I0 is defined as Y×X×I0, where X and Y represent the upsampling matrix and the downsampling matrix respectively.

[0013] Furthermore, in the step 2, two positive normalization coefficients ɑ and β are defined in the energy function to force the energy function to focus on the image edge sharpness L sharpness (s) and texture smoothness L smoothness (s), the energy function is expressed as E(s)=L image (s)+αL sharpness (s)+βL smoothness (s), where L image (s) is the pixel loss, and s is the vector coordinate of the image patch.

[0014] Furthermore, in step 2, the image edge sharpness is expressed as: Texture smoothness is represented by L smoothness (s)=∑||var(I(s+(Δx,Δy)))-var(I(s))||, where s is the vector coordinate of the image block, s includes the local features of the local search and the non-local features obtained by random walk, and I(s) is an image block centered on s. Represents the operator for obtaining the gradient, var represents the operator for obtaining the variance of the image block, and Θ is the feasible domain of the search and matching process.

[0015] Furthermore, the matched image block is represented as I(s+(Δx,Δy)), where (Δx,Δy) is a coordinate offset vector group.

[0016] Furthermore, in step 2, the best matching image block b * Compared with the loss function E(a,b * )=|E(a)-E(b * )| is initialized to a large positive number M.

[0017] Furthermore, in step three, using the formula x i+1 =x i +λ·u i ′ to simulate the movement of particles, where x i is the initial position of the particle, x i+1is the position of the particle at the next moment, and λ is used to control the step size of each movement. λ is initially a large value. After failing to find the optimal value for n consecutive times, its value is updated to λ / 2. u i ′ is the direction of random walk, first randomly generate a two-dimensional direction vector u i ∈(-1,1) 2 , and then normalized it to get During the iterative process, it gradually turns to a refined search and finally converges to the global optimal solution.

[0018] Furthermore, in step 3, during the iteration process, if the loss function value of the image block at the current position of the particle is smaller than the loss value of the best matching image block before the wandering, then the particle position x at this time should be recorded. i is x best , after walking a and b * The loss function

[0019] A computer device comprising:

[0020] processor;

[0021] a memory for storing executable instructions;

[0022] The processor is configured to read the executable instructions from the memory and execute the executable instructions to implement the random walk-based mine image detail enhancement algorithm as described above.

[0023] A computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the processor implements the random walk-based mine image detail enhancement algorithm as described above.

[0024] The beneficial effect of the present invention is that it innovatively introduces the idea of ​​random walk and proposes a new residual feature updating mechanism, which optimizes the detail enhancement effect through global search and image block matching.

[0025] This invention addresses the problems of existing detail enhancement models, which suffer from poor generalization, high complexity, or ineffectiveness. By bypassing the local extreme point search and matching process, the invention finds the optimal solution globally, obtaining the best-matched image to enhance detail enhancement. This significantly improves image detail enhancement, significantly reduces artifacts and halos, and excels in complex mine environments with uneven lighting and high dust content, significantly improving the accuracy of safety monitoring equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The present invention will be further described below with reference to the accompanying drawings and examples.

[0027] Figure 1 This is an algorithm flow chart of the random walk-inspired mine image detail enhancement algorithm in the present invention.

[0028] Figure 2 This is an algorithm structure diagram of the random walk-inspired mine image detail enhancement algorithm in the present invention.

[0029] Figure 3 It is a simulation diagram of random walk in the present invention.

[0030] Figure 4 This is a schematic diagram comparing the processing results of the image detail enhancement algorithm in the present invention and other existing algorithms on the four-fold detail magnification image of the CUMT-CUMID_train_LR_bicubic dataset 0014.

[0031] Figure 5 This is a schematic diagram comparing the processing results of the image detail enhancement algorithm in the present invention and other existing algorithms on the four-fold detail magnification image of the CUMT-CUMID_train_LR_bicubic dataset 0043.

[0032] Figure 6 This is a schematic diagram comparing the processing results of the image detail enhancement algorithm in the present invention and other existing algorithms on the four-fold detail magnification image of image 0413 of the CUMT-CUMID_train_LR_bicubic dataset. DETAILED DESCRIPTION

[0033] The present invention will now be described in further detail with reference to the accompanying drawings, which are simplified schematic diagrams that illustrate the basic structure of the present invention in a schematic manner.

[0034] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", "axial", "radial", "circumferential" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, features defined as "first" or "second" may explicitly or implicitly include one or more of such features. In the description of the present invention, unless otherwise specified, "multiple" means two or more.

[0035] In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed, detachable, or integral connections; mechanical or electrical connections; direct or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on the specific circumstances.

[0036] Example 1

[0037] A random walk-inspired algorithm for enhancing underground mine image details utilizes the randomness and global search capabilities of random walks to escape from local optimal solutions. Specifically, first, the original image I0 is bilinearly interpolated twice to obtain I1 and I2, and the residual Res of the original image I0 is obtained as Res = I0-I2. Secondly, the image block search and matching process is regarded as a process of minimizing the difference (not just pixel differences) of our rewritten energy equation. A random walk algorithm is used for search and matching to obtain the best matching image block, and the enhanced residual f is obtained using our new feature optimization mechanism. new (Res). Finally, we get the detail enhanced image I enhanced =I0+η×f new The present invention significantly enhances image detail and reduces artifacts and halos. It performs well in complex mine environments with uneven lighting and high dust content, significantly improving the accuracy of safety monitoring equipment.

[0038] A random walk-inspired mine image detail enhancement algorithm includes the following steps:

[0039] In the first step, the smooth layer and detail layer (initial residual features) of the original image block are separated by two bilinear interpolations.

[0040] Let the original image be I0, where X and Y represent the upsampling and downsampling matrices, respectively, with their scaling factors being reciprocals of each other. Therefore, the detail layer I1 = X × I0, and the smoothing layer I2 of I0 = Y × X × I0. Since bilinear interpolation is inherently rough, even if the textures of I0 and I2 are very similar, there are still differences. Therefore, the residual Res of the original image I0 is defined as I0 - I2.

[0041] The second step is to initialize the random walk's "starting point" position (original image block), "node" position (image block to be matched), and the loss function of the best matching image block and the original image block.

[0042] S2.1 Initialization of the Random Walk Starting Point When searching for a match for each image block, the starting point of the random walk is set to the center point of the current image block.

[0043] S2.2 New energy equation, as shown in formula (1). The previous way of obtaining residuals, the existing residual acquisition method only uses the characteristics of the local area, and its energy function only focuses on the pixel loss L image (s), as shown in formula (2). With the introduction of random walk algorithm, our method can utilize the global features of the image. In fact, for image detail enhancement, in addition to paying attention to the pixel loss of the image, its edge sharpness L sharpness (s) and texture smoothness L smoothness (s) also needs to be paid attention to. In the energy function we designed, we define two positive normalization coefficients α and β to force the energy function to pay attention to the above issues.

[0044] E(s)=L image (s)+αL sharpness (s)+βL smoothness (s) (1)

[0045]

[0046] In the above formula, s is the vector coordinate of the image block, I(s) is an image block centered on s, represents the gradient operator, var represents the variance operator for the image patch, and Θ is the feasible region for the search and matching process. It is important to note that (Δx, Δy) is a set of coordinate offset vectors, and I(s+(Δx, Δy)) represents the matched image patch. To leverage the speed and performance of the top-level matching mechanism, while also considering the importance of image edges and texture, we define (Δx, Δy) as:

[0047]

[0048] It should be noted that the feasible region s in formula (5) includes not only the local features of the local search, but also the non-local features obtained by random walk, which helps us find the global optimal solution.

[0049] S2.3 Based on the designed new energy function, the best matching image block b * Compared with the loss function E(a,b * )=|E(a)-E(b * )| is initialized to a large positive number M.

[0050] In the third step, the random walk is used to randomly select a walk direction, with the step size controlled by the number of walks. The loss function between the image block to be matched and the original image block and the loss function between the best matching image block and the original image block are calculated. With the goal of minimizing the loss function, the algorithm searches for a better matching image block and continuously updates the walk position.

[0051] S3.1 Escape from Local Optima. Our method escapes local optima by simulating the chaotic motion of particles in space. Previous greedy algorithms would likely stop exploring after finding a local optimum. However, particles' motion in space is unpredictable, so even after finding a local optimum, they will continue exploring, ultimately exploring the entire space.

[0052] In image processing, we use the formula x i+1 =x i +λ·u i ′ to simulate the movement of particles, where x i is the initial position of the particle, x i+1 is the position of the particle at the next moment, and λ is used to control the step size of each movement. λ is initially a large value. After failing to find the optimal value for n consecutive times, its value is updated to λ / 2, gradually turning to a refined search, and finally converging to the global optimal solution. In particular, u i ′ is the direction of random walk, which is obtained by first randomly generating a two-dimensional direction vector u i ∈(-1,1) 2 , and then normalized it to get

[0053] S3.2 During the iteration process, the original image block a and the current best matching image block b * , the original image block a and each image block to be searched b i The energy loss functions of are calculated using formulas (6) and (7) respectively.

[0054] E n (s)=E(a,b * )=|E(a)-E(b * )| (6)

[0055] E n+1 (s)=E(a,b i )=|E(a)-E(b i )| (7)

[0056] S3.3 Update position. If the loss function value (Formula (7)) of the image block at the current position of the particle is smaller than the loss value (Formula (6)) of the best matching image block before the walk, then the particle position x at this time should be recorded. i is x bestOtherwise, the position of the particle should not be recorded. n ′(s) represents a and b after walking * The loss function of , its value can be expressed by formula (8):

[0057]

[0058] In the fourth step, the algorithm iterates through a certain number of random walks, gradually converging to the globally optimal matching image patch and completing the update of the residual features. The updated residual is then amplified and superimposed with the input original image to obtain the final detail-enhanced image.

[0059] S4.1 Iterative Update. This process repeats step 2 until the convergence condition is met, that is, the step size λ is less than the convergence factor ε. λ and ε are parameters that need to be manually adjusted to control the convergence accuracy.

[0060] S4.2 Update of residual features. We use the random walk-based residual update mechanism f new (·), using the initial residual features obtained in the first step, calculate the enhanced detail features f new (Res).

[0061] S4.3 obtains the final detail-enhanced image I enhanced =I0+η×f new (Res), where I0 is the original image and η is the detail magnification factor that needs to be manually adjusted to obtain better visual effects.

[0062] The following is the pseudo code of the random walk-based detail enhancement algorithm:

[0063]

[0064]

[0065] Returns: Position x best The best matching image block b corresponding to

[0066] Note: E(a,b)=|E(a)-E(b)|

[0067] Next, the final implementation effect of the present invention will be further explained in conjunction with specific experiments. Currently, the more effective image detail enhancement algorithms include GIF, GGIF, WGIF, BFLS, TH, ZF, ILS and IPRH. The data sets used in the experiment include classic data sets such as Set14, General100, BSD100, and the CUMT-CUMID_train_LR_bicubic data set. The CUMT-CUMID data set is an image set collected from multiple real mines based on professional mining cameras. The image samples therein are real and complete records and reflect the impact of different environments and lighting conditions in the mine on the quality of the images collected by the equipment, so that the image samples in the data set are widely applicable to the design, development and testing of various image processing algorithms for mine environments.

[0068] In order to quantitatively verify the effectiveness of the proposed algorithm, this embodiment provides a numerical comparison of objective evaluation indicators with other algorithms. The evaluation indicators we use are the internationally commonly used PSNR (Peak Signal-to-Noise Ratio) and SSIM (Structure Similarity Index Measure). PSNR is an error-sensitive image quality evaluation method used to measure the difference between two images. The larger the PSNR value, the smaller the difference between the two images, that is, the better the corresponding algorithm; SSIM is an image quality evaluation indicator that measures image similarity from three aspects: brightness, contrast, and structure. The value range of SSIM is [0,1]. The closer its value is to 1, the smaller the image distortion is, and the better the corresponding algorithm is.

[0069]

[0070] The calculation methods of PSNR and SSIM are shown in Equations (9) and (10); MSE (Mean Square Error) is the mean square error between the original image I and the image K generated by the specific algorithm, h×w is the resolution of images I and K; μ I and μ K The mean of the original image and the image generated by the specific algorithm, and The variance of the original image and the image generated by the specific algorithm, σ IK The covariance between the original image and the image generated by a specific algorithm; C1 and C2 are constants set to avoid the denominator being zero.

[0071] Based on the above calculation method, this example provides quantitative comparison results of the proposed algorithm and other mainstream algorithms (including GIF, GGIF, WGIF, BFLS, TH, ZF, and ILS) on the Set14, General100, and BSD100 datasets, as shown in Table 1:

[0072] Table 1 Quantitative comparison results

[0073]

[0074] Referring to Table 1, the algorithm proposed in this embodiment performs best in all SSIM tests, indicating that the present invention can most perfectly protect the image structure information while enhancing the image details. At the same time, the algorithm proposed in this embodiment is optimal or suboptimal in the PSNR test, indicating that the proposed algorithm can effectively preserve the image details while reducing noise and distortion, thereby better maintaining the image clarity and quality.

[0075] This embodiment also provides an internationally accepted Mean Opinion Score (MOS) to measure the subjective visual performance of each algorithm. The specific process is to select a certain number of experts and volunteers without relevant professional knowledge, and randomly select the same number of pictures from each data set, anonymously display the images processed by each algorithm to them, and give a certain score based on each person's naked eye judgment of the image quality; after removing extreme scores, calculate the average score of each image and sort them in order of high and low. The results of the MOS score are shown in Table 2. Our algorithm ranked first or second in the MOS test, which shows that the RW algorithm has strong robustness, reliable performance in subjective vision, improves the intuitive visual comfort of the vast majority of people, and can effectively enhance the details of various images. In particular, it performs better in images under mines, and is very suitable for application in mine environments.

[0076] Table 2 MOS score ranking of various algorithms

[0077]

[0078] In order to illustrate the effect of the present invention from the actual visual performance, Figure 4 、 Figure 5 、 Figure 6 The visual effects of the algorithm proposed in this embodiment are compared with those of some mainstream algorithms on the CUMT-CUMID_train_LR_bicubic dataset. Specifically, Figure 4The a-picture in the figure is the CUMT-CUMID_train_LR_bicubic dataset 0014 image, a1-picture is the original image with four times the details in the red box in a-picture, a2-picture is the image processed by the BFLS algorithm on a1-picture, and its values ​​in the numerical evaluation of PSNR and SSIM are 16.89dB and 0.5475 respectively, a3-picture is the image processed by the GIF algorithm on a1-picture, and its values ​​in the numerical evaluation of PSNR and SSIM are 19.18dB and 0.6437 respectively, a4-picture is the image processed by the TH algorithm on a1-picture, and its values ​​in the numerical evaluation of PSNR and SSIM are 18.36dB and 0.6123 respectively, Small figure 5 is the image after small figure a1 is processed by the algorithm RW in this application, and its values ​​in the numerical evaluation of PSNR and SSIM are 25.12dB and 0.8692 respectively. Small figure a6 is the image after small figure a1 is processed by the GGIF algorithm, and its values ​​in the numerical evaluation of PSNR and SSIM are 16.82dB and 0.5777 respectively. Small figure a7 is the image after small figure a1 is processed by the ILS algorithm, and its values ​​in the numerical evaluation of PSNR and SSIM are 18.43dB and 0.5799 respectively. Small figure a8 is the image after small figure a1 is processed by the IPRH algorithm, and its values ​​in the numerical evaluation of PSNR and SSIM are 23.29dB and 0.8224 respectively. Figure 5The b-picture in the figure is the CUMT-CUMID_train_LR_bicubic dataset 0043 image, the b1-picture is the four-fold detail magnification of the original image in the red box in the b-picture, the b2-picture is the image after the b1-picture is processed by the BFLS algorithm, and its values ​​in the numerical evaluation of PSNR and SSIM are 15.52dB and 0.5447 respectively, the b3-picture is the image after the b1-picture is processed by the GIF algorithm, and its values ​​in the numerical evaluation of PSNR and SSIM are 17.68dB and 0.6274 respectively, the b4-picture is the image after the b1-picture is processed by the TH algorithm, and its values ​​in the numerical evaluation of PSNR and SSIM are 18.36dB and 0.6123 respectively, Figure 5 is the image after image b1 is processed by the algorithm RW in this application. Its values ​​in the numerical evaluation of PSNR and SSIM are 23.07dB and 0.8451 respectively. Figure b6 is the image after image b1 is processed by the GGIF algorithm. Its values ​​in the numerical evaluation of PSNR and SSIM are 15.00dB and 0.5471 respectively. Figure b7 is the image after image b1 is processed by the ILS algorithm. Its values ​​in the numerical evaluation of PSNR and SSIM are 16.52dB and 0.5852 respectively. Figure b8 is the image after image b1 is processed by the IPRH algorithm. Its values ​​in the numerical evaluation of PSNR and SSIM are 21.17dB and 0.8011 respectively. Figure 6The c-th image is the CUMT-CUMID_train_LR_bicubic dataset 0413 image, the c1-th image is the four-fold magnified original image of the red box in the c-th image, the c2-th image is the image processed by the BFLS algorithm on the c1-th image, and its values ​​in the numerical evaluation of PSNR and SSIM are 16.80dB and 0.5090 respectively, the c3-th image is the image processed by the GIF algorithm on the c1-th image, and its values ​​in the numerical evaluation of PSNR and SSIM are 19.44dB and 0.6137 respectively, the c4-th image is the image processed by the TH algorithm on the c1-th image, and its values ​​in the numerical evaluation of PSNR and SSIM are 19.09dB and 0.5886 respectively, Figure 5 is the image after image c1 is processed using the RW algorithm in this application. Its values ​​in the numerical evaluation of PSNR and SSIM are 28.13dB and 0.8848 respectively. Figure c6 is the image after image c1 is processed using the GGIF algorithm. Its values ​​in the numerical evaluation of PSNR and SSIM are 17.36dB and 0.5664 respectively. Figure c7 is the image after image c1 is processed using the ILS algorithm. Its values ​​in the numerical evaluation of PSNR and SSIM are 18.48dB and 0.5617 respectively. Figure c8 is the image after image c1 is processed using the IPRH algorithm. Its values ​​in the numerical evaluation of PSNR and SSIM are 25.77dB and 0.8391 respectively.

[0079] As can be easily seen from the figure, other algorithms suffer from varying degrees of overfitting, incorrect image color restoration, noise amplification, and halo effects, leading to image distortion. However, the algorithm proposed in this embodiment effectively enhances image detail while maintaining a natural and harmonious overall effect and minimizing distortion. Therefore, the algorithm proposed in this embodiment has significant significance and practical value in mine image analysis and the modernization of mine construction.

[0080] In summary, the experimental results above demonstrate the effectiveness and advantages of the present invention. The present invention adopts the idea of ​​random walk to avoid falling into local optimum and searches for image blocks in a global range, thereby achieving better image detail enhancement effect.

[0081] Example 2

[0082] An embodiment of the present application provides a computer device, which includes a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or at least one program is loaded and executed by the processor to implement a random walk-based mine image detail enhancement algorithm provided in the above-mentioned method embodiment.

[0083] Example 3

[0084] The embodiment of the present application also provides a computer-readable storage medium, which can be set in a server to store at least one instruction or at least one program for implementing a random walk mine image detail enhancement algorithm in the method embodiment, and the at least one instruction or the at least one program is loaded and executed by the processor to implement the random walk-based mine image detail enhancement algorithm provided in the above method embodiment. Optionally, in this embodiment, the above-mentioned storage medium can be located in at least one network server among multiple network servers in a computer network. Optionally, in this embodiment, the above-mentioned storage medium can include but is not limited to: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program code.

[0085] With the above-described preferred embodiments of the present invention as a guide, and with reference to the above description, relevant personnel are fully capable of making various changes and modifications without departing from the technical spirit of this invention. The technical scope of this invention is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. A random walk-based mine image detail enhancement algorithm, characterized in that: include, Step 1: obtain the smooth layer I2 and detail layer I1 of the original image I0 through two bilinear interpolation separations, and calculate the initial residual feature Res; Step 2: Initialize the "starting point" position of the random walk, which is the original image block, and the "node" position, which is the image block to be matched, and calculate the loss function of the best matching image block and the original image block; Step 3: With the goal of minimizing the loss function, find a better matching image block and continuously update the wandering position; Step 4: Through continuous iteration of random walks, the algorithm gradually converges to the globally optimal matching image block, completes the update of the residual features, amplifies the updated residual and superimposes it with the input original image to obtain the final detail-enhanced image.

2. The random walk-based mine image detail enhancement algorithm according to claim 1, characterized in that: In the step 1, the residual Res of the original image I0 is defined as I0-I2, and the smoothing layer I2 of the original image I0 is defined as Y×X×I0, where X and Y represent the upsampling matrix and the downsampling matrix respectively.

3. The random walk-based mine image detail enhancement algorithm according to claim 1, characterized in that: In the second step, two positive normalization coefficients ɑ and β are defined in the energy function to force the energy function to focus on the image edge sharpness L sharpness (s) and texture smoothness L smoothness (s), the energy function is expressed as E(s)=L image (s)+αL sharpness (s)+βL smoothness (s), where L image (s) is the pixel loss, and s is the vector coordinate of the image patch.

4. The random walk-based mine image detail enhancement algorithm according to claim 3, characterized in that: In step 2, the image edge sharpness is expressed as: Texture smoothness is represented by L smoothness (s)=∑||var(I(s+(Δx,Δy)))-var(I(s))||, where s is the vector coordinate of the image block, s includes the local features of the local search and the non-local features obtained by random walk, and I(s) is an image block centered on s. Represents the operator for obtaining the gradient, var represents the operator for obtaining the variance of the image block, and Θ is the feasible domain of the search and matching process.

5. The random walk-based mine image detail enhancement algorithm according to claim 4, characterized in that: The matched image block is represented as I(s+(Δx,Δy)), where (Δx,Δy) is a coordinate offset vector group.

6. The random walk-based mine image detail enhancement algorithm according to claim 5, characterized in that: In step 2, the best matching image block b * Compared with the loss function E(a,b * )=|E(a)-E(b * )| is initialized to a large positive number M.

7. The random walk-based mine image detail enhancement algorithm according to claim 1, characterized in that: In step 3, use the formula x i+1 =x i +λ·u i ′ to simulate the movement of particles, where x i is the initial position of the particle, x i+1 is the position of the particle at the next moment, and λ is used to control the step size of each movement. λ is initially a large value. After failing to find the optimal value for n consecutive times, its value is updated to λ / 2. u i ′ is the direction of random walk, first randomly generate a two-dimensional direction vector u i ∈(-1,1) 2 , and then normalized it to get During the iterative process, it gradually turns to a refined search and finally converges to the global optimal solution.

8. The random walk-based mine image detail enhancement algorithm according to claim 7, characterized in that: In step 3, during the iteration process, if the loss function value of the image block at the current position of the particle is smaller than the loss value of the best matching image block before the walk, then the particle position x at this time should be recorded. i is x best , after walking a and b * The loss function 9. A computer device, characterized in that: include: processor; a memory for storing executable instructions; The processor is configured to read the executable instructions from the memory and execute the executable instructions to implement the random walk-based mine image detail enhancement algorithm according to any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor implements the random walk-based underground mine image detail enhancement algorithm according to any one of claims 1 to 8.

Images