Image denoising method and system

Abstract

The present application relates to a kind of image denoising method and system, belong to image processing technical field, solve the problem of image blur caused by the introduction of artificial artifact in the process of image denoising in prior art.The image denoising method includes: for each pixel point in the image to be denoised, determine a plurality of search regions according to the preset neighborhood of pixel point;Determine the denoising image corresponding to each search region by step S21 and step S22;Image fusion is carried out on the denoising image corresponding to a plurality of search regions, and the denoising image after image fusion is used as the final denoising image.The clarity of the image after denoising is improved.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to an image noise reduction method and system. Background Art

[0002] Since images are inevitably contaminated by noise to varying degrees during the process of acquisition, transmission, and display, which reduces image quality, the main goal of image denoising is to filter out random noise and retain image details as much as possible.

[0003] Nonlocal means (NLM) is one of the mainstream methods for image denoising and has attracted widespread attention in the field. Originally designed to eliminate Gaussian noise, this method leverages the global self-similarity of an image to achieve image denoising. Self-similarity is evaluated based on the grayscale differences between pixel neighborhoods (i.e., image patches) rather than individual pixels. In recent years, researchers have extended NLM to denoise images contaminated by speckle noise (such as medical ultrasound images and synthetic aperture radar images), proposing probabilistic patch-based (PPB) filtering methods and optimized Bayesian nonlocal means (OBNLM) methods, which have achieved some success.

[0004] However, in this denoising scheme, the calculation of the weights is affected by the noise itself, which introduces artificial artifacts in the image denoising process, causing the image to become blurred and reducing the quality of the denoised image. Summary of the Invention

[0005] In view of the above analysis, the embodiments of the present invention aim to provide an image denoising method and system to solve the problem of image blurring caused by artificial artifacts introduced during the image denoising process in the prior art.

[0006] In one aspect, an embodiment of the present invention provides an image denoising method, the image denoising method comprising:

[0007] Acquire the image to be denoised, perform edge filling on the image to be denoised, and obtain an initial image to be denoised; for each pixel in the denoised image, determine multiple search areas based on a preset neighborhood of the pixel;

[0008] The denoised image corresponding to each search area is determined by the following steps:

[0009] Step S21: Using the NL-means algorithm, the weights of each pixel corresponding to the previous denoised image are obtained as the weights of each pixel in the current denoising process; and the current denoised image is obtained based on the weights of each pixel in the current denoising process and the initial image to be denoised.

[0010] Step S22: Determine whether the weights of each pixel in the current denoising process and the weights of each pixel in the previous denoising process meet a preset iteration stop condition; if not, use the current denoised image as the previous denoising image and jump to step S21; otherwise, use the current denoised image as the denoised image corresponding to the search area;

[0011] The denoised images corresponding to the various search areas are fused, and the denoised image after image fusion is used as the final denoised image.

[0012] Based on the further improvement of the above method, the multiple search areas are determined according to the preset neighborhood of the pixel point, including:

[0013] Determine a plurality of pixel positions included in a preset neighborhood, and use each pixel position as the center of each search area to determine a plurality of search areas;

[0014] The size of the preset neighborhood is (2S+1)*(2S+1), the size of each search area is (2S'+1)*(2S'+1), and S'>S, and both S and S' are positive integers.

[0015] Based on a further improvement of the above method, the image fusion of the noise reduction images corresponding to the multiple search areas includes:

[0016] The average value of the pixel values ​​of each pixel in the denoised image corresponding to the multiple search areas is calculated as the pixel value of each pixel in the final denoised image.

[0017] Based on a further improvement of the above method, the determination of whether the weights of each pixel of the current denoising and the weights of each pixel of the previous denoising meet a preset iteration stop condition includes:

[0018] Traverse the weights of each pixel in this denoising and the previous denoising respectively, and determine the average value and variance of the weights corresponding to each pixel in this denoising and the previous denoising;

[0019] Construct a value function for each pixel. When the value function of each pixel meets the preset convergence threshold, the preset iteration stop condition is met; otherwise, it is not met.

[0020] Based on the further improvement of the above method, the value function of each pixel is as follows:

[0021]

[0022] in, Represents the value function of the i-th pixel in this denoising, Represents the average value of the weight corresponding to the i-th pixel in this denoising. Represents the average value of the weight corresponding to the i-th pixel in the last denoising; Indicates the variance of the weight corresponding to the i-th pixel in this denoising. Indicates the variance of the weight corresponding to the i-th pixel in the last denoising.

[0023] Based on the further improvement of the above method, when the value function of each pixel satisfies the following formula, the preset iteration stop condition is met:

[0024]

[0025] Where L represents the total number of pixels, and δ represents the preset convergence threshold.

[0026] Based on the further improvement of the above method, the NL-means algorithm is used to obtain the weights of each pixel corresponding to the last denoised image, including:

[0027]

[0028]

[0029] Among them, w(i,j) represents the jth weight corresponding to the i-th pixel, v(N i ) represents the neighborhood centered on the i-th pixel, v(N j ) represents the neighborhood centered on the j-th pixel, represents the Gaussian weighted Euclidean distance between two neighborhoods based on grayscale, a represents the standard deviation of the Gaussian kernel, h represents the filtering parameter, Z(i) represents the normalization coefficient, and Q represents the number of neighbors corresponding to the neighborhood centered on the i-th pixel in this search area.

[0030] Based on a further improvement of the above method, the denoised image is obtained according to the weights of the pixels to be denoised and the initial image to be denoised, including:

[0031] The pixel value corresponding to each pixel of the denoised image is calculated using the following formula:

[0032]

[0033] Wherein, NL(v)(i) represents the pixel value of the i-th pixel of the current denoised image, and v(j) represents the pixel value of the j-th pixel of the initial denoised image.

[0034] Based on the further improvement of the above method, the edge filling of the denoised image can be performed in any of the following ways:

[0035] Reproduction method;

[0036] Reflection method;

[0037] outer packaging law;

[0038] Constant method;

[0039] Furthermore, the size of the initial image to be denoised obtained after edge filling is (m+2S)*(n+2S); wherein the size of the image to be denoised before edge filling is m*n.

[0040] In another aspect, an embodiment of the present invention provides an image noise reduction system, comprising:

[0041] A filling module is used to obtain the image to be denoised, perform edge filling on the denoised image to obtain an initial image to be denoised; for each pixel in the denoised image, determine multiple search areas based on a preset neighborhood of the pixel;

[0042] The denoising module is used to determine the denoised image corresponding to each search area through the following steps:

[0043] Step S21: Using the NL-means algorithm, the weights of each pixel corresponding to the previous denoised image are obtained as the weights of each pixel in the current denoising process; and the current denoised image is obtained based on the weights of each pixel in the current denoising process and the initial image to be denoised.

[0044] Step S22: Determine whether the weights of each pixel in the current denoising process and the weights of each pixel in the previous denoising process meet a preset iteration stop condition; if not, use the current denoised image as the previous denoising image and jump to step S21; otherwise, use the current denoised image as the denoised image corresponding to the search area;

[0045] The fusion module is used to fuse the denoised images corresponding to the multiple search areas and use the denoised images after image fusion as the final denoised image.

[0046] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects:

[0047] 1. The present invention uses a preset neighborhood, search area and the last denoised image, and a preset denoising algorithm to obtain the weights of each pixel corresponding to the last denoised image, as the weights of each pixel in this denoising, and continuously iteratively updates the weights of each pixel; at the same time, the initial denoised image is used to obtain the denoised image, so that the denoised image has higher clarity and better quality.

[0048] 2. The present invention determines multiple search areas through preset neighborhoods of pixel points, and performs image fusion on the noise reduction images corresponding to the multiple search areas, so that the final noise reduction image is smoother and more natural.

[0049] In the present invention, the above-mentioned technical solutions can be combined with each other to achieve more preferred combinations. Other features and advantages of the present invention will be described in the following description, and some advantages will become apparent from the description or be learned through practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the contents particularly pointed out in the description and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The accompanying drawings are only for the purpose of illustrating particular embodiments and are not to be considered limiting of the present invention. Like reference symbols denote like parts throughout the drawings.

[0051] Figure 1 A schematic diagram of a flow chart of an image noise reduction method provided by an embodiment of the present invention;

[0052] Figure 2 A schematic diagram of the structure of a preset neighborhood provided in an embodiment of the present invention;

[0053] Figure 3(a) shows the Figure 2 A schematic diagram of a search area structure in which pixel position 4 is the center of the search area;

[0054] Figure 3(b) shows the Figure 2 A schematic diagram of a search area structure in which pixel position 5 is the center of the search area;

[0055] Figure 3(c) shows the Figure 2 A schematic diagram of a search area structure in which the pixel position 6 is the center of the search area;

[0056] Figure 4 Schematic diagram of the structure for calculating the jth weight corresponding to the i-th pixel;

[0057] Figure 5 This is a structural diagram of an image noise reduction system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0058] The preferred embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, and are not used to limit the scope of the present invention.

[0059] A specific embodiment of the present invention discloses an image noise reduction method, such as Figure 1 As shown, the image noise reduction method includes:

[0060] Step S1: Acquire the image to be denoised, perform edge filling on the image to be denoised, and obtain an initial image to be denoised; for each pixel in the denoised image, determine multiple search areas based on a preset neighborhood of the pixel;

[0061] Step S2: Determine the noise reduction image corresponding to each search area through the following steps:

[0062] Step S21: Using the NL-means algorithm, the weights of each pixel corresponding to the previous denoised image are obtained as the weights of each pixel in the current denoising process; and the current denoised image is obtained based on the weights of each pixel in the current denoising process and the initial image to be denoised.

[0063] Step S22: Determine whether the weights of each pixel in the current denoising process and the weights of each pixel in the previous denoising process meet a preset iteration stop condition; if not, use the current denoised image as the previous denoising image and jump to step S21; otherwise, use the current denoised image as the denoised image corresponding to the search area;

[0064] Step S3: performing image fusion on the denoised images corresponding to the various search areas, and using the fused denoised image as the final denoised image.

[0065] Specifically, such as Figure 1 As shown, in step S1, edge filling is performed on the image to be denoised. It is understandable that after edge filling, the edges of the initial image to be denoised will be clearer and more continuous. In addition, the pixels at the very edge cannot be processed during the image denoising process. Therefore, by performing edge filling on the image to be denoised, the pixel count of the image after denoising can be increased.

[0066] Preferably, the edge filling of the denoised image is performed in any of the following ways:

[0067] Reproduction method;

[0068] Reflection method;

[0069] outer packaging law;

[0070] Constant method;

[0071] Furthermore, the size of the initial image to be denoised obtained after edge filling is (m+2S)*(n+2S); wherein the size of the image to be denoised before edge filling is m*n.

[0072] Specifically, the copy method can copy the pixels at the edge of the image to be denoised, and then fill them to the periphery of the edge of the image to be denoised to obtain the initial image to be denoised.

[0073] Specifically, the reflection method can use a layer of pixels at the edge of the image to be denoised as a symmetry axis, and fill the pixels inside the symmetry axis to the periphery of the symmetry axis to obtain an initial image to be denoised.

[0074] Specifically, the outer packaging method can fill the pixels on the left edge of the image to be denoised to the periphery of the right edge of the image to be denoised, and at the same time fill the pixels on the right edge to the periphery of the left edge of the image to be denoised; similarly, fill the pixels on the upper edge to the periphery of the lower edge of the image to be denoised, and fill the pixels on the lower edge to the periphery of the upper edge of the image to be denoised, to obtain the initial image to be denoised.

[0075] Specifically, in the constant method, pixels with a fixed pixel value are filled into the periphery of the edge of the image to be denoised to obtain an initial image to be denoised.

[0076] Specifically, the size of the image to be denoised before edge filling is m*n, that is, the width of the image to be denoised is set to m and the height is set to n, and the size of the initial image to be denoised after edge filling is (m+2S)*(n+2S), that is, the width of the initial image to be denoised after edge filling is (m+2S)*(n+2S), where the size of the preset neighborhood is (2S+1)*(2S+1). It can be understood that the width of the preset neighborhood is (2S+1) and the height is (2S+1).

[0077] It is worth noting that when filling the edges, the image to be denoised is filled according to the size of a preset neighborhood, which can improve computational efficiency.

[0078] Specifically, such as Figure 1 As shown, in step S1, when performing noise reduction on the image to be denoised, a neighborhood is set in advance, which is recorded as a preset neighborhood.

[0079] Preferably, the determining of multiple search areas based on preset neighborhoods of pixel points includes:

[0080] Determine a plurality of pixel positions included in a preset neighborhood, and use each pixel position as the center of each search area to determine a plurality of search areas;

[0081] The size of the preset neighborhood is (2S+1)*(2S+1), the size of each search area is (2S'+1)*(2S'+1), and S'>S, and both S and S' are positive integers.

[0082] Specific, exemplary, such as Figure 2 As shown, when S=1, the size of the preset neighborhood is 3*3, and the preset neighborhood includes 9 pixel positions, including pixel position 1, pixel position 2... and pixel position 9. At this time, S' can be 2, and the size of the search area is 5*5.

[0083] It is worth noting that, according to the 9 pixel positions included in the preset neighborhood, 9 pixel positions can be selected and the 9 pixel positions can be used as the centers of the search areas to determine 9 types of search areas.

[0084] It can be understood that, according to the 9 pixel positions included in the preset neighborhood, pixel positions that are symmetrical to the center pixel position 5 of the neighborhood area can be selected, such as selecting 4 pixel positions of pixel position 2, pixel position 4, pixel position 6 and pixel position 8, and using pixel position 2, pixel position 4, pixel position 6 and pixel position 8 as the centers of the search areas respectively to determine 4 search areas; at the same time, according to the 9 pixel positions included in the preset neighborhood, pixel positions that are symmetrical to the center pixel position 5 of the neighborhood area can be selected, such as selecting 4 pixel positions of pixel position 1, pixel position 3, pixel position 7 and pixel position 9, and using pixel position 1, pixel position 3, pixel position 7 and pixel position 9 as the centers of the search areas respectively to determine 4 search areas.

[0085] Specifically, as shown in FIG3(a), FIG3(b) and FIG3(c), Figure 2 Schematic diagram of the structure of the search area in which pixel position 4, pixel position 5 and pixel position 6 are the centers of the search area.

[0086] It is worth noting that when the size of the preset neighborhood is 3*3, 9 pixel positions can be selected to determine 9 search areas respectively; or 4 pixel positions can be selected to determine 4 search areas respectively.

[0087] Specifically, when the size of the preset neighborhood is (2S+1)*(2S+1), the size of each search area is (2S'+1)*(2S'+1), and S'>S, and S and S' are both positive integers.

[0088] Specifically, such as Figure 1 As shown, in step S2, the noise reduction image corresponding to each search area is determined through steps S21 and S22.

[0089] Specifically, in step S21, preferably, the weight of each pixel corresponding to the last denoised image is obtained using the NL-means algorithm, including:

[0090]

[0091]

[0092] Among them, w(i,j) represents the jth weight corresponding to the i-th pixel, v(N i ) represents the neighborhood centered on the i-th pixel, v(N j) represents the neighborhood centered on the j-th pixel, represents the Gaussian weighted Euclidean distance between two neighborhoods based on grayscale, a represents the standard deviation of the Gaussian kernel, h represents the filtering parameter, Z(i) represents the normalization coefficient, and Q represents the number of neighbors corresponding to the neighborhood centered on the i-th pixel in this search area.

[0093] It can be understood that when denoising is performed for the first time, the last denoised image refers to the initial image to be denoised.

[0094] Specifically, the non-local means (NL-means) algorithm is a new denoising technique proposed in recent years. This method fully utilizes redundant information in the image to maximize image detail while denoising. The basic idea is that the estimated value of the current pixel is the weighted average of pixels in the image with similar neighborhood structures.

[0095] Specifically, when a search area is determined with the pixel position 5 of the preset neighborhood as the center of the search area, such as Figure 4 As shown, the weight corresponding to pixel i is to be calculated, with pixel i as the center of the preset neighborhood (assuming S=1, the preset neighborhood size is 3*3, S' is 3, the size of the search area is 7*7, and the sliding step is 1). It can be understood that pixel j can move 4 times to the right and 4 times downward according to the sliding step 1, and it can be determined that there are 25 neighborhoods corresponding to pixel i with pixel j as the neighborhood center. It is worth noting that when the preset neighborhood size is 3*3 and the size of the search area is 7*7, there are 25 neighborhoods corresponding to each pixel i whose weight is to be calculated, that is, the Q value is 25. Therefore, the j-th weight corresponding to pixel i can be calculated by the following formula:

[0096]

[0097]

[0098] Preferably, obtaining the current denoised image according to the weights of the pixels to be denoised and the initial image to be denoised includes:

[0099] The pixel value corresponding to each pixel of the denoised image is calculated using the following formula:

[0100]

[0101] Wherein, NL(v)(i) represents the pixel value of the i-th pixel of the current denoised image, and v(j) represents the pixel value of the j-th pixel of the initial denoised image.

[0102] Specifically, after calculating all weight values ​​corresponding to the i-th pixel, the pixel value corresponding to the i-th pixel can be calculated using the following formula:

[0103]

[0104] Specifically, in step S22, it is determined whether the weights of each pixel of the current denoising and the weights of each pixel of the previous denoising meet the preset iteration stop condition; if not, the current denoised image is used as the previous denoised image and the process jumps to step S21; otherwise, the current denoised image is used as the denoised image corresponding to the search area.

[0105] Preferably, the determining whether the weights of the pixels subjected to current denoising and the weights of the pixels subjected to previous denoising satisfy a preset iterative stop condition includes:

[0106] Traverse the weights of each pixel in this denoising and the previous denoising respectively, and determine the average value and variance of the weights corresponding to each pixel in this denoising and the previous denoising;

[0107] Construct a value function for each pixel. When the value function of each pixel meets the preset convergence threshold, the preset iteration stop condition is met; otherwise, it is not met.

[0108] Specifically, such as Figure 4 As shown, after calculating the 25 weights corresponding to the i-th pixel, the average value and variance of the weights corresponding to the i-th pixel in the current denoising and the previous denoising can be determined respectively.

[0109] Specifically, such as Figure 1 As shown, in step S22, a value function of each pixel is constructed. When the value function of each pixel meets the preset convergence threshold, the preset iteration stop condition is met; otherwise, it is not met.

[0110] Preferably, the value function of each pixel is the following formula:

[0111]

[0112] in, Represents the value function of the i-th pixel in this denoising, Represents the average value of the weight corresponding to the i-th pixel in this denoising. Represents the average value of the weight corresponding to the i-th pixel in the last denoising; Indicates the variance of the weight corresponding to the i-th pixel in this denoising. Indicates the variance of the weight corresponding to the i-th pixel in the last denoising.

[0113] Preferably, the preset iteration stopping condition is met when the value function of each pixel satisfies the following formula:

[0114]

[0115] Where L represents the total number of pixels, and δ represents the preset convergence threshold.

[0116] Specifically, such as Figure 1 As shown, in step S3, the denoised images corresponding to the various search areas are fused, and the denoised image after image fusion is used as the final denoised image.

[0117] Preferably, the image fusion of the noise reduction images corresponding to the multiple search areas includes:

[0118] The average value of the pixel values ​​of each pixel in the denoised image corresponding to the multiple search areas is calculated as the pixel value of each pixel in the final denoised image.

[0119] Specifically, the denoised images corresponding to the various search areas can be calculated through step S2. In step S3, it is determined whether the denoised images corresponding to the various search areas have been calculated. When the calculations are completed, the average value of the pixel values ​​of each pixel in the denoised images corresponding to the various search areas is calculated as the pixel value of each pixel in the final denoised image.

[0120] It is worth noting that when calculating the search area corresponding to each search area in step S2, it can be executed sequentially or in parallel, and there is no specific limitation here. The denoised images corresponding to the search areas calculated in step S2 are respectively transmitted to step S3. In step S3, when it is determined that the denoised images corresponding to all search areas are calculated, the denoised images corresponding to all search areas are fused to obtain the final denoised image.

[0121] The present invention sets multiple search areas through preset neighborhoods of pixel points, fuses the denoised images under each search area, and rationally utilizes multiple pixel positions around the pixel points, avoiding the regional mutation caused by a single search area, making the final denoised image more natural and smooth.

[0122] Another specific embodiment of the present invention discloses an image noise reduction system, such as Figure 5 As shown, the image noise reduction system includes:

[0123] The filling module 100 is used to obtain the image to be denoised, perform edge filling on the image to be denoised, and obtain an initial image to be denoised; for each pixel in the denoised image, determine multiple search areas based on a preset neighborhood of the pixel;

[0124] The noise reduction module 200 is configured to determine the noise reduction image corresponding to each search area through the following steps:

[0125] Step S21: Using the NL-means algorithm, the weights of each pixel corresponding to the previous denoised image are obtained as the weights of each pixel in the current denoising process; and the current denoised image is obtained based on the weights of each pixel in the current denoising process and the initial image to be denoised.

[0126] Step S22: Determine whether the weights of each pixel in the current denoising process and the weights of each pixel in the previous denoising process meet a preset iteration stop condition; if not, use the current denoised image as the previous denoising image and jump to step S21; otherwise, use the current denoised image as the denoised image corresponding to the search area;

[0127] The fusion module 300 is used to fuse the denoised images corresponding to the various search areas, and use the fused denoised image as the final denoised image.

[0128] Compared with the existing technology, the image denoising method and system provided in this embodiment uses a preset neighborhood, search area and the previous denoised image, and a preset denoising algorithm to obtain the weights of each pixel corresponding to the previous denoised image, as the weights of each pixel for this denoising, and continuously iteratively updates the weights of each pixel; at the same time, the initial image to be denoised is used to obtain the current denoised image, so that the denoised image has higher clarity and better quality; in addition, multiple search areas are determined through the preset neighborhood of pixel points, and the denoised images corresponding to the multiple search areas are fused, so that the final denoised image is smoother and more natural.

[0129] Those skilled in the art will appreciate that all or part of the process steps of the above-described embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, such as a magnetic disk, an optical disk, a read-only memory, or a random access memory.

[0130] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by any technician familiar with this technical field within the technical scope disclosed by the present invention should be covered by the scope of protection of the present invention.

Claims

1. An image denoising method, characterized in that: The image noise reduction method comprises: Acquire the image to be denoised, perform edge filling on the image to be denoised, and obtain an initial image to be denoised; for each pixel in the denoised image, determine multiple search areas based on a preset neighborhood of the pixel; The denoised image corresponding to each search area is determined by the following steps: Step S21: Using the NL-means algorithm, the weights of each pixel corresponding to the previous denoised image are obtained as the weights of each pixel in the current denoising process; and the current denoised image is obtained based on the weights of each pixel in the current denoising process and the initial image to be denoised. Step S22: Determine whether the weights of each pixel in the current denoising process and the weights of each pixel in the previous denoising process meet a preset iteration stop condition; if not, use the current denoised image as the previous denoising image and jump to step S21; otherwise, use the current denoised image as the denoised image corresponding to the search area; Perform image fusion on the denoised images corresponding to the multiple search areas, and use the fused denoised image as the final denoised image; The determining whether the weights of the pixels in the current denoising and the weights of the pixels in the previous denoising satisfy a preset iteration stop condition includes: Traverse the weights of each pixel in this denoising and the previous denoising respectively, and determine the average value and variance of the weights corresponding to each pixel in this denoising and the previous denoising; Construct the value function of each pixel. When the value function of each pixel meets the preset convergence threshold, the preset iteration stop condition is met; otherwise, it is not met; The value function of each pixel is as follows: in, Represents the value function of the i-th pixel in this denoising, Represents the average value of the weight corresponding to the i-th pixel in this denoising. Represents the average value of the weight corresponding to the i-th pixel in the last denoising; Indicates the variance of the weight corresponding to the i-th pixel in this denoising. Indicates the variance of the weight corresponding to the i-th pixel in the last denoising.

2. The image denoising method according to claim 1, wherein: The method of determining multiple search areas based on preset neighborhoods of pixel points includes: Determine a plurality of pixel positions included in a preset neighborhood, and use each pixel position as the center of each search area to determine a plurality of search areas; The size of the preset neighborhood is (2S+1)*(2S+1), the size of each search area is (2S'+1)*(2S'+1), and S'>S, and both S and S' are positive integers.

3. The image denoising method according to claim 2, wherein: The image fusion of the noise reduction images corresponding to the multiple search areas includes: The average value of the pixel values ​​of each pixel in the denoised image corresponding to the multiple search areas is calculated as the pixel value of each pixel in the final denoised image.

4. The image denoising method according to claim 1, wherein: When the value function of each pixel satisfies the following formula, the preset iteration stop condition is met: Where L represents the total number of pixels, and δ represents the preset convergence threshold.

5. The image denoising method according to claim 2, wherein: Use the NL-means algorithm to obtain the weights of each pixel corresponding to the last denoised image, including: Among them, w(i,j) represents the jth weight corresponding to the i-th pixel, v(N i ) represents the neighborhood centered on the i-th pixel, v(N j ) represents the neighborhood centered on the j-th pixel, represents the Gaussian weighted Euclidean distance between two neighborhoods based on grayscale, a represents the standard deviation of the Gaussian kernel, h represents the filtering parameter, Z(i) represents the normalization coefficient, and Q represents the number of neighbors corresponding to the neighborhood centered on the i-th pixel in this search area.

6. The image denoising method according to claim 5, wherein: The step of obtaining the denoised image based on the weights of the pixels to be denoised and the initial image to be denoised includes: The pixel value corresponding to each pixel of the denoised image is calculated using the following formula: Wherein, NL(v)(i) represents the pixel value of the i-th pixel of the current denoised image, and v(j) represents the pixel value of the j-th pixel of the initial denoised image.

7. The image denoising method according to claim 2, wherein: Fill the edges of the denoised image using any of the following methods: Reproduction method; Reflection method; outer packaging law; Constant method; Furthermore, the size of the initial image to be denoised obtained after edge filling is (m+2S)*(n+2S); wherein the size of the image to be denoised before edge filling is m*n.

8. An image noise reduction system, characterized in that: The image noise reduction system comprises: A filling module is used to obtain the image to be denoised, perform edge filling on the denoised image to obtain an initial image to be denoised; for each pixel in the denoised image, determine multiple search areas based on a preset neighborhood of the pixel; The denoising module is used to determine the denoised image corresponding to each search area through the following steps: Step S21: Using the NL-means algorithm, the weights of each pixel corresponding to the previous denoised image are obtained as the weights of each pixel in the current denoising process; and the current denoised image is obtained based on the weights of each pixel in the current denoising process and the initial image to be denoised. Step S22: Determine whether the weights of each pixel in the current denoising process and the weights of each pixel in the previous denoising process meet a preset iteration stop condition; if not, use the current denoised image as the previous denoising image and jump to step S21; otherwise, use the current denoised image as the denoised image corresponding to the search area; A fusion module is used to fuse the denoised images corresponding to the multiple search areas and use the fused denoised images as the final denoised image; The determining whether the weights of the pixels in the current denoising and the weights of the pixels in the previous denoising satisfy a preset iteration stop condition includes: Traverse the weights of each pixel in this denoising and the previous denoising respectively, and determine the average value and variance of the weights corresponding to each pixel in this denoising and the previous denoising; Construct the value function of each pixel. When the value function of each pixel meets the preset convergence threshold, the preset iteration stop condition is met; otherwise, it is not met; The value function of each pixel is as follows: in, Represents the value function of the i-th pixel in this denoising, Represents the average value of the weight corresponding to the i-th pixel in this denoising. Represents the average value of the weight corresponding to the i-th pixel in the last denoising; Indicates the variance of the weight corresponding to the i-th pixel in this denoising. Indicates the variance of the weight corresponding to the i-th pixel in the last denoising.

Images