Image denoising method, system, and storage medium

Abstract

The application discloses an image denoising method, system and storage medium, and relates to the technical field of image processing, and comprises the following steps: performing fuzzy processing on an image to be processed through empirical mode decomposition to obtain a fuzzy feature image; performing edge detection processing on the feature image to obtain an edge detection operator of the feature image; calculating a diffusion threshold of a preset anisotropic diffusion equation according to the edge detection operator, and determining an improved anisotropic diffusion equation according to the calculated diffusion threshold; and performing diffusion processing on the image to be processed through the improved anisotropic diffusion equation to obtain image information after denoising. The above method can weaken the influence of image noise on the edge detection result and improve the image denoising effect.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to an image denoising method, system and storage medium. Background Art

[0002] Image denoising is a relatively important branch of image processing. A common approach is to obtain a denoised image by solving the anisotropic diffusion equation on the original image. This approach can remove noise while maintaining a certain degree of features. The image's gradient operator determines the diffusion rate and is widely used in image denoising, image segmentation, and target recognition. However, the gradient operator itself is highly sensitive to noise and has poor noise immunity, making it unable to distinguish pseudo-edges caused by noise. To address the gradient operator's poor noise recognition capabilities, edge detection operators were later introduced to improve this approach. However, noise can create edges that don't exist, leading to inaccurate edge detection results. This makes image denoising using the anisotropic diffusion equation ineffective. Summary of the Invention

[0003] The present application aims to solve at least one of the technical problems existing in the prior art. To this end, the present application proposes an image denoising method, system and storage medium that can reduce the impact of image noise on edge detection results and improve the image denoising effect.

[0004] In a first aspect, the present application proposes an image denoising method, comprising:

[0005] The image to be processed is fuzzy processed by empirical mode decomposition to obtain a fuzzy feature image;

[0006] Performing edge detection processing on the feature image to obtain an edge detection operator of the feature image;

[0007] Calculating a preset diffusion threshold of anisotropic diffusion equation according to the edge detection operator, and determining an improved anisotropic diffusion equation according to the calculated diffusion threshold;

[0008] The image to be processed is subjected to diffusion processing by using the improved anisotropic diffusion equation to obtain denoised image information.

[0009] The image denoising method according to the embodiment of the first aspect of the present application has at least the following beneficial effects: by blurring the image before edge detection, the high-frequency part corresponding to the details of the image is removed, and the noise attached to the detail part of the image is eliminated at this time. Then, by performing edge detection on the blurred image, the noise in the detail part cannot generate edges by itself, so that the edge detection result is accurate, and the edge detection operator in the edge detection process is not affected by noise, thereby replacing the diffusion threshold in the anisotropic diffusion equation to accurately guide the diffusion degree, thereby outputting a denoised image with better denoising effect.

[0010] According to some embodiments of the present application, the image to be processed is blurred by empirical mode decomposition to obtain a blurred feature image, including

[0011] Performing two-dimensional empirical mode decomposition on the image to be processed to obtain an intrinsic mode function set of the image to be processed; the intrinsic mode function set consists of an image residual and a plurality of image function components arranged from high to low frequency;

[0012] The intrinsic mode function set is reorganized by removing the image function component with the highest frequency to obtain the characteristic image.

[0013] According to some embodiments of the present application, the two-dimensional empirical mode decomposition is performed on the image to be processed to obtain the intrinsic mode function set of the image to be processed, including

[0014] determining an image function according to the image to be processed;

[0015] Calculating image function components corresponding to the image function according to a preset decomposition formula;

[0016] Calculating a difference between the image function and the image function component;

[0017] Taking the difference that does not meet the preset condition as a new image function and recalculating the image function component corresponding to the new image function until the difference meets the preset condition;

[0018] The difference obtained by the last calculation is used as the image residual;

[0019] The image residual and each calculated image function component are combined to obtain the intrinsic mode function set.

[0020] According to some embodiments of the present application, calculating the image function components corresponding to the image function according to a preset decomposition formula includes:

[0021] According to the maximum value and the minimum value of the image function, constructing a maximum value surface and a minimum value surface respectively;

[0022] Calculating a mean envelope surface based on the maximum value surface and the minimum value surface;

[0023] Calculating the difference between the mean envelope surface and the image function to obtain a height difference function;

[0024] The height difference function is used as a new image function and a new height difference function is recalculated until the mean square error between the new height difference function and the height difference function calculated the previous time satisfies a preset threshold.

[0025] According to some embodiments of the present application, the edge detection processing is performed on the feature image to obtain the edge detection operator of the feature image information, including:

[0026] Performing Gaussian smoothing on the feature image to obtain smoothed image information;

[0027] Performing gradient calculation on the smoothed image information to obtain a gradient magnitude and gradient direction for each pixel in the smoothed image;

[0028] According to the gradient direction, non-maximum suppression processing is performed on the corresponding gradient amplitude to obtain optimized image information;

[0029] Applying double-threshold screening to the optimized image information to obtain screened image information;

[0030] According to the filtered image information, the edge detection operator and the corresponding edge distribution are determined.

[0031] According to some embodiments of the present application, determining an edge detection operator and an edge distribution based on the filtered image information includes:

[0032] performing a hysteresis edge tracking process on the filtered image information to obtain a valid edge set;

[0033] According to the effective edge set, an edge operator and its corresponding edge distribution are determined.

[0034] According to some embodiments of the present application, the diffusion coefficient of the improved anisotropic diffusion equation represents the relationship between the pixel points of the image to be processed and the gradient operator.

[0035] According to some embodiments of the present application, the image to be processed is diffused using the improved anisotropic diffusion equation to obtain denoised image information, including:

[0036] Obtaining the preset number of calculations of the improved anisotropic diffusion equation;

[0037] Performing diffusion processing on the image to be processed by using the improved anisotropic diffusion equation, and outputting a first diffusion image;

[0038] Using the diffusion image as the new image to be processed, and re-performing diffusion processing until the number of diffusion processing times matches a preset number of calculation times or the diffusion image meets a preset condition;

[0039] The diffused image obtained by the last diffusion process is used as the denoised image information.

[0040] In a second aspect, the present application proposes an image denoising system, comprising: at least one memory, at least one processor, and at least one program, wherein the program is stored in the memory, and the processor executes at least one program to implement the image denoising method as described in any one of claims 1 to 8.

[0041] According to the image denoising system of the second aspect embodiment of the present application, there are at least the following beneficial effects: by blurring the image before edge detection, the high-frequency part corresponding to the details of the image is removed, and the noise attached to the detail part of the image is eliminated at this time. Then, by performing edge detection on the blurred image, the noise in the detail part cannot generate edges by itself, so that the edge detection result is accurate and the edge detection operator in the edge detection process is not affected by noise, thereby replacing the diffusion threshold in the anisotropic diffusion equation to accurately guide the diffusion degree, thereby outputting a denoised image with better denoising effect.

[0042] In a third aspect, the present application proposes a computer-readable storage medium, wherein the computer-readable storage medium stores a computer-executable signal, and the computer-executable signal is used to execute the image denoising method according to any one of claims 1 to 8.

[0043] According to the computer-readable storage medium of the third aspect embodiment of the present application, there are at least the following beneficial effects: by blurring the image before edge detection, the high-frequency part corresponding to the details of the image is removed, and the noise attached to the detail part of the image is eliminated at this time. Then, by performing edge detection on the blurred image, the noise in the detail part cannot generate edges by itself, so that the edge detection result is accurate, and the edge detection operator in the edge detection process is not affected by noise, thereby replacing the diffusion threshold in the anisotropic diffusion equation to accurately guide the diffusion degree, thereby outputting a denoised image with better denoising effect.

[0044] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become obvious from the description below, or will be learned through practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Additional aspects and advantages of the present application will become more apparent from the following detailed description, taken in conjunction with the accompanying drawings, illustrating by way of example the principles of the application.

[0046] Figure 1 is a flow chart of an image denoising method provided by one embodiment of the present application;

[0047] Figure 2 is a detailed flow chart of step S100 in Figure 1

[0048] Figure 3 is a detailed flow chart of step S110 in Figure 2

[0049] Figure 4 is a detailed flow chart of step S112 in Figure 3

[0050] Figure 5 is a detailed flow chart of step S200 in Figure 1

[0051] Figure 6 is a detailed flow chart of step S250 in Figure 5

[0052] Figure 7 is a detailed flow chart of step S400 in Figure 1

[0053] Figure 8 is a structure diagram of an image denoising system of the present application. DETAILED DESCRIPTION

[0054] In order to make the objects, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and should not be used to limit the present application.

[0055] In the description of the present application, one or more is meant one or more, more than two is meant more than two, greater than, less than, more than, etc. are understood as not including the number, above, below, etc. are understood as including the number. If it is described as first, second, it is only used to distinguish technical features for the purpose, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features or implicitly indicating the sequence of indicated technical features.

[0056] Referring to Figure 1 , in a first aspect, the present application provides an image denoising method, including but not limited to the following steps: ​​​​​​

[0057] Step S100, blurring the image to be processed by empirical mode decomposition to obtain a blurred feature image.

[0058] In some embodiments, the image to be processed is divided into several different components according to frequency by two-dimensional empirical mode decomposition, and the components required in the subsequent steps are extracted as needed. The image is blurred by removing the components related to the image details from the above components and retaining the basic outline of the image, thereby obtaining the blurred feature image information.

[0059] Step S200, edge detection processing is performed on the feature image to obtain an edge detection operator of the feature image.

[0060] In some embodiments, the blurred feature image obtained above only includes the basic features and outlines of the image, and the noise part coexisting with the image details has been removed. By performing edge detection on the feature image and selecting a reasonable edge detection operator to guide the detection process, an edge distribution functionally related to the edge detection operator is obtained, which is used to improve the anisotropic diffusion equation in the subsequent steps.

[0061] Step S300, calculating the diffusion threshold of the preset anisotropic diffusion equation according to the edge detection operator, and determining the improved anisotropic diffusion equation according to the calculated diffusion threshold.

[0062] It can be understood that the edge detection operator contains the function information of the edge distribution. By replacing the gradient operator of the anisotropic diffusion equation with the edge detection operator, the diffusion speed and degree of the anisotropic diffusion equation are guided, thereby improving the anisotropic diffusion equation.

[0063] In some embodiments, the original anisotropic diffusion equation has the expression,

[0064]

[0065] where I(x, y, 0) is the initial image, i.e., the image to be processed, div represents the divergence operator, and represents the gradient operator; c represents the diffusion coefficient, which controls the diffusion speed of P-M (anisotropic diffusion equation). The edge distribution obtained by the edge detection process guided by the canny operator is substituted into the anisotropic diffusion equation, thereby generating a new diffusion threshold with the gradient operator, guiding the new diffusion equation to complete the denoising process. The improved anisotropic diffusion equation is:

[0066]

[0067] Here, K represents the canny operator. In the edge area, K approaches 1, the speed of the diffusion equation slows down, and the detail information is retained; in the non-edge area, K approaches 0, the diffusion speed is accelerated, and the denoising effect is achieved.

[0068] Step S400 , performing diffusion processing on the image to be processed by using the improved anisotropic diffusion equation to obtain denoised image information.

[0069] In some embodiments, by performing image degradation processing before edge detection, the high-frequency parts corresponding to the image details are removed, thereby eliminating the noise attached to the image details. Then, by performing edge detection on the degraded image, the noise in the details cannot generate edges on its own, so that the edge detection result is accurate and the edge detection operator in the edge detection process is not affected by noise. This replaces the diffusion threshold in the anisotropic diffusion equation to accurately guide the diffusion degree, thereby outputting a denoised image with better denoising effect.

[0070] According to the embodiment of the first aspect of the present application, at least the following beneficial effects are achieved: by blurring the image before edge detection, the high-frequency part corresponding to the details of the image is removed, thereby eliminating the noise attached to the detail part of the image. Then, by performing edge detection on the blurred image, the noise in the detail part cannot generate edges on its own, so that the edge detection result is accurate, and the edge detection operator in the edge detection process is not affected by noise, thereby replacing the diffusion threshold in the anisotropic diffusion equation to accurately guide the diffusion degree, thereby outputting a denoised image with better denoising effect.

[0071] It is understandable that, referring to Figure 2 ,in Figure 2 yes Figure 1 FIG. 1 is a schematic diagram of an embodiment of a detailed process of step S100 in FIG. 1 , wherein step S100 includes:

[0072] Step S110 , performing two-dimensional empirical mode decomposition on the image to be processed to obtain an intrinsic mode function set of the image to be processed.

[0073] It should be noted that for an unknown signal, empirical mode decomposition can be started directly without prior analysis and research. This method automatically divides the signal into layers according to some fixed modes without the need for manual settings and intervention. The decomposed intrinsic mode function is based on the characteristics of the signal itself, and subsequent research can be conducted on specific frequency information. Since the analysis object of this application is an image, a two-dimensional empirical mode decomposition for image analysis is used.

[0074] In step S120, the intrinsic mode function set of the image function component with the highest frequency removed is reassembled to obtain a feature image.

[0075] In some embodiments, by selecting and recombining the decomposed intrinsic mode functions, the intrinsic mode function components that affect the subsequent edge detection accuracy are removed, and the remaining intrinsic mode functions are recombined, the blurring of the image is accurately controlled to avoid uncontrollable blurring, resulting in too little blurring and too much noise, and too much blurring and loss of information required for subsequent edge detection.

[0076] It should be noted that step S100 includes but is not limited to step S110 and step S120.

[0077] It is understandable that, referring to Figure 3 ,in Figure 3 yes Figure 2 Schematic diagram of an embodiment of the detailed process of step S110 in FIG. 1 , wherein step S110 includes but is not limited to step S111 , step S112 , step S113 , step S114 , step S115 and step S116 .

[0078] Step S111: determining an image function according to the image to be processed.

[0079] In some embodiments, it is assumed that the image to be processed is f(x, y), which is a function of image pixels.

[0080] Step S112: Calculate the image function components corresponding to the image function according to the preset decomposition formula.

[0081] Step S113, calculating the difference between the image function and the image function components.

[0082] In some embodiments, the difference between the image function obtained by some decomposition rules and the image function generated by the previous decomposition process is calculated, and the difference is used as the new image function. At the same time, when the difference of each cycle meets the characteristic preset conditions, the difference of that time will be used as an intrinsic mode function in the two-dimensional empirical mode decomposition process.

[0083] Step S114 : taking the difference that does not meet the preset condition as a new image function and recalculating the image function components corresponding to the new image function until the difference meets the preset condition.

[0084] Step S115: The difference obtained by the last calculation is used as the image residual.

[0085] It can be understood that after multiple cycles of step S114, multiple new image functions are generated, namely the differences mentioned above. At the same time, there are cases where the differences cannot be used as new image functions. Let r1(x,y) be the image remaining after the first image function component is removed from the image f(x,y). Using r1(x,y) as the new image function input, repeat steps S111, S112 and S113 to obtain the second difference. Let the first image function component be C1(x,y), then r1(x,y)=f(x,y)-C1(x,y). When the above steps are repeated n times, r n (x,y) and C n (x,y) is less than the predetermined error or r n When (x, y) is a monotonic function, the image function component can no longer be extracted from the image f(x, y). n (x,y) represents the average trend of the image slowly changing, and r n (x,y) is used as the image residual for the last calculation.

[0086] Step S116: Combine the image residual and each calculated image function component to obtain an intrinsic mode function set.

[0087] In some embodiments, the original image f(x,y) can be expressed as:

[0088]

[0089] It should be noted that the feature image in step S120 can be expressed as:

[0090]

[0091] It is understandable that, referring to Figure 4 ,in Figure 4 yes Figure 3 FIG. 5 is a schematic diagram of an embodiment of a detailed process of step S112 in FIG. 5 , wherein step S112 includes but is not limited to step S500 , step S510 , step S520 and step S530 .

[0092] Step S500: construct a maximum surface and a minimum surface according to the maximum and minimum values ​​of the image function.

[0093] In some embodiments, the maximum and minimum values ​​in the image f(x, y) are first found, and a maximum surface u(x, y) and a minimum surface v(x, y) are constructed by interpolation.

[0094] Step S510: Calculate the mean envelope surface based on the maximum surface and the minimum surface.

[0095] In some embodiments, by solving the mean formula of the maximum surface and the minimum surface, the specific expression is:

[0096]

[0097] Step S520: Calculate the difference between the mean envelope surface and the image function to obtain a height difference function.

[0098] In some embodiments, the difference between the mean e1(x,y) and f(x,y) is denoted as h1(x,y), then

[0099] h1(x,y)=f(x,y)-e1(x,y)

[0100] Step S530: The height difference function is used as a new image function and a new height difference function is recalculated until the mean square error between the new height difference function and the height difference function calculated previously meets a preset threshold.

[0101] In some embodiments, h1(x, y) is used instead of f(x, y) as a new input, and steps S500 to S513 are repeated. Assuming that the steps are repeated K times in total, the threshold value that satisfies the condition for stopping the above-mentioned repetition steps is calculated as follows:

[0102]

[0103] Where SD represents the threshold value, which is set between 0.1 and 0.5. When the above steps are repeated K times and the value of SD meets the specified range, the loop of steps S500 to S513 is stopped. 1k (x, y) is an intrinsic mode function (IMF), that is, the image function component mentioned in the above step S114, which is the image function component with the highest frequency of the image f(x, y), denoted as c1(x, y).

[0104] It is understandable that, referring to Figure 5 ,in Figure 5 yes Figure 1 Schematic diagram of an embodiment of the detailed process of step S200 in FIG. 1 , wherein step S200 includes but is not limited to step S210 , step S220 , step S230 , step S240 and step S250 .

[0105] Step S210: Perform Gaussian smoothing on the feature image to obtain smoothed image information.

[0106] In some embodiments, the feature image obtained by the two-dimensional empirical mode decomposition is filtered and smoothed with a Gaussian filter to perform preliminary image pre-noise reduction, so that the edges of the feature image are smoother, thereby avoiding large errors in subsequent edge detection results.

[0107] Step S220: performing gradient calculation on the smoothed image information to obtain the gradient magnitude and gradient direction of each pixel in the smoothed image.

[0108] In some embodiments, the gradient is the difference between two adjacent pixel values. The difference in the X and Y directions needs to be calculated, and finally the gradient G and gradient direction θ of the pixel point can be obtained, where Simplified to G = |Gx| + |Gy|; θ = arctan(Gy / Gx).

[0109] In step S230, non-maximum suppression is performed on the corresponding gradient amplitude according to the gradient direction to obtain optimized image information.

[0110] In some embodiments, non-maximum suppression converts "fat" edges into "thin edges." Grayscale changes are concentrated in the local area, and the largest grayscale changes in the gradient direction within the local area are retained, while the others are not retained. This can turn a multi-pixel edge into a single-pixel wide edge.

[0111] In step S240, the optimized image information is subjected to dual-threshold screening to obtain screened image information.

[0112] In some embodiments, an upper threshold and a lower threshold are generally set. When the edge pixel value is greater than the upper threshold, it is recorded as a strong edge; when the edge pixel value is greater than the lower threshold and less than the upper threshold, it is recorded as a weak edge; when the edge pixel value is less than the lower threshold, it is rejected. This application automatically feeds back the upper and lower thresholds through library functions. If the pixel points are distributed from 0 to 1, the upper threshold is 0.1563 and the lower threshold is 0.0625.

[0113] Step S250: Determine the edge detection operator and the corresponding edge distribution based on the filtered image information.

[0114] It is understandable that, through the processing of the above steps, the edge distribution of the feature image input for edge detection is expressed in the form of an edge detection operator, thereby completing edge detection. In some embodiments, the present application adopts a canny detection operator.

[0115] It is understandable that, referring to Figure 6 ,in Figure 6 yes Figure 5 Schematic diagram of an embodiment of the detailed process of step S250 in FIG. 1 , wherein step S250 includes but is not limited to step S251 and step S252.

[0116] Step S251: Perform hysteresis edge tracking processing on the filtered image information to obtain a valid edge set.

[0117] It is understandable that the edge detection process described above classifies edge types into strong and weak edges. Strong edges can be considered true edges, while weak edges can be either true edges or false edges caused by noise or color shifts. It is generally believed that weak edges caused by true edges are connected to strong edge points, while weak edge points caused by noise are not, thus requiring lagged edge tracking. By detecting pixels in the 8-connected area of ​​a weak edge, as long as a strong edge point exists, the weak edge is considered true and retained, thus distinguishing a finite set of edges.

[0118] Step S252: Determine the edge operator and its corresponding edge distribution based on the valid edge set.

[0119] It will be appreciated that the diffusion coefficient of the improved anisotropic diffusion equation represents the relationship between pixels in the image to be processed and the gradient operator. In some embodiments, the diffusion coefficient represents the diffusion velocity of the anisotropic diffusion equation. The edge detection operator guides the edge detection process, and the output edge distribution is a function of the edge detection operator. This function, combined with the diffusion coefficient, forms a new diffusion threshold, thereby deriving the improved anisotropic diffusion equation.

[0120] It is understandable that, referring to Figure 7 ,in Figure 7 yes Figure 1 Schematic diagram of an embodiment of the detailed process of step S400 in FIG. 4 , wherein step S400 includes but is not limited to step S410 , step S420 , step S430 and step S440 .

[0121] Step S410: Obtain the number of calculations of the improved anisotropic diffusion equation.

[0122] In some embodiments, the preset number of initial calculations in this application is 5. It is understandable that a small number of cycles can be performed first to observe whether the output image initially meets the denoising requirements before determining the subsequent output conditions.

[0123] Step S420: performing diffusion processing on the image to be processed by using an improved anisotropic diffusion equation, and outputting a first diffusion image.

[0124] Step S430: Using the diffusion image as the new image to be processed, the diffusion process is performed again until the number of diffusion processes matches the preset number of calculations or the diffusion image meets a preset condition.

[0125] Step S440: Using the diffused image obtained by the last diffusion process as denoised image information.

[0126] In some embodiments, the peak signal-to-noise ratio (PSNR) of the original image is compared with the peak signal-to-noise ratio (PSNR) of the denoised image. When the PSNR of the denoised image is greater than that of the original image, it indicates that the denoised image is relatively ideal and can be output as the final denoised image. The peak signal-to-noise ratio (PSNR) is used as an indicator for analysis. It is a considerable image evaluation indicator based on the error between corresponding pixels. The larger the PSNR, the smaller the degree of distortion, and the closer it is to the true image. The calculation formula is as follows:

[0127]

[0128] Wherein, f2(x, y) is the image diffused by the improved anisotropic diffusion equation after blurring, f(x, y) is the image to be processed, and M and N are the sizes of the corresponding images. At this time, the PSNR value can be used as one of the loop conditions (preset conditions) for stopping the diffusion process. When the PSNR of the diffused image output in step S420 meets the preset PSNR, the diffusion process can be stopped.

[0129] In a second aspect, the present application also provides an image optimization system, comprising at least one memory, at least one processor and at least one program, wherein the program is stored in the memory, and the processor executes one or more programs to implement the above-mentioned image denoising method.

[0130] This image optimization system uses an optimization method to blur the image before edge detection, removing high-frequency components corresponding to image details and eliminating noise associated with these details. Edge detection is then performed on the blurred image, preventing the noise in the details from generating edges on their own. This ensures accurate edge detection results and unaffected edge detection operators during the edge detection process. This replaces the diffusion threshold in the anisotropic diffusion equation to accurately guide the diffusion level, resulting in a denoised image with improved denoising performance.

[0131] The processor and memory can be connected via a bus or other means. Figure 8 Take a processor as an example, connected via a bus.

[0132] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer-executable programs, and signals, such as the program instructions / signals corresponding to the processing module in the embodiments of the present application. The processor executes the non-transitory software programs, instructions, and signals stored in the memory to perform various functional applications and data processing, thereby implementing the image denoising method of the above-mentioned method embodiment.

[0133] The memory can include a program storage area and a data storage area, wherein the program storage area can store an operating system, application programs required by at least one function; and the data storage area can store related data of the above-mentioned vehicle driving steering safety early warning method, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state memory device. In some embodiments, the memory can optionally include a memory disposed remotely with respect to the processor, and these remote memories can be connected to the processing module through a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0134] One or more signals are stored in the memory, and when executed by one or more processors, perform the image denoising method in any of the above-mentioned method embodiments. For example, the method steps S100 to S400 in the above-mentioned method embodiment, Figure 1 the method steps S110 to S120 in the above-mentioned method embodiment, Figure 2 the method steps S111 to S116 in the above-mentioned method embodiment, Figure 3 the method steps S500 to S530 of the above-mentioned method embodiment, Figure 4 the steps S210 to S250 in the above-mentioned method embodiment, Figure 5 the steps S251 to S252 in the above-mentioned method embodiment, and Figure 6 the steps S410 to S440 in the above-mentioned method embodiment. Figure 7

[0135] In a third aspect, the embodiments of the present application provide a computer readable storage medium, which stores computer executable instructions, and the computer executable instructions are executed by one or more processors, so that the above-mentioned one or more processors perform the image denoising method in the above-mentioned method embodiments. For example, the method steps S100 to S400 in the above-mentioned method embodiment, Figure 1 the method steps S110 to S120 in the above-mentioned method embodiment, Figure 2 the method steps S111 to S116 in the above-mentioned method embodiment, Figure 3 the method steps S500 to S530 of the above-mentioned method embodiment, Figure 4 the steps S210 to S250 in the above-mentioned method embodiment, Figure 5 the steps S251 to S252 in the above-mentioned method embodiment, and Figure 6 the steps S410 to S440 in the above-mentioned method embodiment. Figure 7

[0136] ​​The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and 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 may be selected based on actual needs to achieve the objectives of this embodiment.

[0137] Through the description of the above embodiments, it will be appreciated by those skilled in the art that all or some of the steps and systems in the method disclosed above can be implemented as software, firmware, hardware and appropriate combinations thereof. Some physical components or all physical components can be implemented as software executed by a processor, such as a central processing unit, a digital signal processor or a microprocessor, or implemented as hardware, or implemented as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, and the computer-readable medium can include a computer storage medium (or non-transitory medium) and a communication medium (or temporary medium). As known to those skilled in the art, the term computer storage medium is included in any method or technology for storing information (such as a computer-readable signal, a data structure, a program module or other data) and is volatile and non-volatile, removable and non-removable. Computer storage media includes but is not limited to RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cassette, magnetic tape, disk storage or other magnetic storage device, or any other medium that can be used to store desired information and can be accessed by a computer. Furthermore, as is well known to those skilled in the art, communication media typically embodies computer-readable signals, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

[0138] The embodiments of the present application are described in detail above in conjunction with the accompanying drawings, but the present application is not limited to the above embodiments. Various changes can be made within the scope of knowledge possessed by ordinary technicians in the relevant technical field without departing from the purpose of the present application.

Claims

1. An image denoising method, characterized in that: include: Performing fuzzy processing on the image to be processed by empirical mode decomposition to obtain a fuzzy feature image, performing two-dimensional empirical mode decomposition on the image to be processed to obtain an intrinsic mode function set of the image to be processed, the intrinsic mode function set consisting of an image residual and a plurality of image function components arranged from high to low frequency, and recombining the intrinsic mode function set after removing the image function component with the highest frequency to obtain the feature image; Performing edge detection processing on the feature image to obtain an edge detection operator of the feature image; Calculating a preset diffusion threshold of anisotropic diffusion equation according to the edge detection operator, and determining an improved anisotropic diffusion equation according to the calculated diffusion threshold; The image to be processed is subjected to diffusion processing by using the improved anisotropic diffusion equation to obtain denoised image information.

2. The image denoising method according to claim 1, wherein: The two-dimensional empirical mode decomposition is performed on the image to be processed to obtain the intrinsic mode function set of the image to be processed, including determining an image function according to the image to be processed; Calculating image function components corresponding to the image function according to a preset decomposition formula; Calculating a difference between the image function and the image function component; Taking the difference that does not meet the preset condition as a new image function and recalculating the image function component corresponding to the new image function until the difference meets the preset condition; The difference obtained by the last calculation is used as the image residual; The image residual and each calculated image function component are combined to obtain the intrinsic mode function set.

3. The image denoising method according to claim 2, wherein: The step of calculating the image function components corresponding to the image function according to a preset decomposition formula includes: According to the maximum value and the minimum value of the image function, constructing a maximum value surface and a minimum value surface respectively; Calculating a mean envelope surface based on the maximum value surface and the minimum value surface; Calculating the difference between the mean envelope surface and the image function to obtain a height difference function; The height difference function is used as a new image function and a new height difference function is recalculated until the mean square error between the new height difference function and the height difference function calculated the previous time satisfies a preset threshold.

4. The image denoising method according to claim 1, wherein: The edge detection process is performed on the feature image to obtain the edge detection operator of the feature image, including Performing Gaussian smoothing on the feature image to obtain smoothed image information; Performing gradient calculation on the smoothed image information to obtain a gradient magnitude and gradient direction for each pixel in the smoothed image; According to the gradient direction, non-maximum suppression processing is performed on the corresponding gradient amplitude to obtain optimized image information; Applying double-threshold screening to the optimized image information to obtain screened image information; According to the filtered image information, the edge detection operator and the corresponding edge distribution are determined.

5. The image denoising method according to claim 4, wherein: Determining the edge detection operator and edge distribution based on the filtered image information includes: performing a hysteresis edge tracking process on the filtered image information to obtain a valid edge set; According to the effective edge set, an edge operator and its corresponding edge distribution are determined.

6. The image denoising method according to claim 1, wherein: The diffusion coefficient of the improved anisotropic diffusion equation represents the relationship between the pixel points of the image to be processed and the gradient operator.

7. The image denoising method according to claim 1, wherein: The improved anisotropic diffusion equation is used to perform diffusion processing on the image to be processed to obtain denoised image information, including Obtaining the number of calculations of the improved anisotropic diffusion equation; Performing diffusion processing on the image to be processed by using the improved anisotropic diffusion equation, and outputting a first diffusion image; Using the diffusion image as the new image to be processed, and re-performing diffusion processing until the number of diffusion processing times matches a preset number of calculation times or the diffusion image meets a preset condition; The diffused image obtained by the last diffusion process is used as the denoised image information.

8. An image denoising system, characterized in that: include: at least one memory; at least one processor; at least one program; The programs are stored in the memory, and the processor executes at least one of the programs to implement the image denoising method according to any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer-executable signal, and the computer-executable signal is used to execute the image denoising method according to any one of claims 1 to 7.

Images