Noise reduction method and apparatus, and electronic device and computer-readable storage medium

Abstract

Provided in the present application are a noise reduction method and apparatus, and an electronic device and a computer-readable storage medium. The method comprises: filtering residual images of an image to generate a filtered image, wherein the residual images are the differences of the image at different resolutions, the image corresponds to at least two residual images, and there is a preset hierarchical sequence relationship between the at least two residual images; enhancing weak edge regions in the filtered image and weakening strong edge regions in the filtered image, so as to generate a sharpened image; and obtaining a target image on the basis of a reconstructed image and the sharpened image, wherein the reconstructed image is the superposition of all residual images in a hierarchical sequence preceding the residual images. During a superposition process of reconstruction, sharpening is performed to increase local contrast and enhance edges and details, without the need for sharpening after image reconstruction; and strong edge regions are weakened, such that noise in the strong edge regions can be reduced and image artifacts can be mitigated.

Description

Noise reduction method, device, electronic device, and computer-readable storage medium Technical Field

[0001] The present application relates to the field of image processing, and more specifically, to a noise reduction method, device, electronic device, and computer-readable storage medium. Background Art

[0002] In the application scenarios of image sensors such as CMOS Sensor (Complementary Metal-Oxide-Semiconductor Sensor) and CCD Sensor (Charge-Coupled Device Sensor) in mobile phones and video surveillance, it is usually necessary to perform spatial noise reduction on the image to improve the signal-to-noise ratio of the output image. However, after noise reduction, the image will become blurry and the image clarity will be relatively low, and subsequent sharpening algorithms are usually required to improve the image clarity. In addition, the noise reduction algorithms commonly used have edge-preserving properties, and the noise reduction strength at strong edges of the image is often insufficient. More noise will remain at the strong edges of the image, resulting in a "burr" feeling at the edges.

[0003] Summary of the Invention

[0004] In view of this, the purpose of the present application is to provide a noise reduction method, device, electronic device and computer-readable storage medium to solve the shortcoming of poor clarity of the noise reduction image generated by the existing noise reduction method.

[0005] In a first aspect, the present application provides a noise reduction method, which includes: filtering a residual image of an image to generate a filtered image, wherein the residual image is the difference between the image at different resolutions, the image corresponds to at least two residual images, and there is a preset hierarchical relationship between the at least two residual images; enhancing the weak edge area in the filtered image and weakening the strong edge area in the filtered image to generate a sharpened image; obtaining a target image based on a reconstructed image and the sharpened image, wherein the reconstructed image is the superposition of all residual images in the hierarchical order before the residual image.

[0006] In the above embodiment, sharpening is performed during the reconstruction process of the image to be denoised, which increases the local contrast of the image and enhances the edges and details in the image. There is no need to sharpen the image after reconstruction, which makes the implementation process simpler and more convenient. In addition, the strong edge area is weakened, which can reduce the noise in the strong edge area and reduce the burrs in the image.

[0007] In an optional embodiment of the present application, obtaining a target image based on the reconstructed image and the sharpened image includes: establishing a guidance map and a noise map based on the reconstructed image and the sharpened image; denoising the noise map based on the guidance map to obtain a denoised image; and upsampling the denoised image to obtain the target image.

[0008] In the above embodiment, the guidance map and the noise map can be established quickly and easily. The use of the guidance map can give more weight to the strong edge area, thereby enhancing the noise reduction strength at the strong edges in the image. The noise map is the result of superimposing the residual image after sharpening, which increases the local contrast of the image, enhances the edges and details in the image, and can make the generated image clearer and sharper.

[0009] In an optional embodiment of the present application, the method further includes: assigning a value to the filtered image based on a preset threshold interval to obtain the weak edge area; assigning a value to the filtered image based on a preset threshold to obtain the strong edge area.

[0010] In the above embodiment, by assigning values ​​using preset thresholds, the strong edge region and the weak edge region of the image can be quickly and simply obtained.

[0011] In an optional embodiment of the present application, the sharpened image includes a weak edge area image, and the enhancing the weak edge area in the filtered image includes: enhancing the weak edge area using a preset first scalar, and retaining other parts outside the weak edge area in the filtered image to generate the weak edge area image.

[0012] In the above embodiment, the weak edge region can be quickly enhanced by using a scalar, while other parts outside the weak edge region are retained, so that the generated image of the weak edge region has more image details.

[0013] In an optional embodiment of the present application, the sharpened image includes a strong edge area image, and the weakening of the strong edge area in the filtered image includes: using a preset second scalar to weaken the strong edge area, and retaining other parts outside the strong edge area in the filtered image to generate the strong edge area image.

[0014] In the above embodiment, the strong edge region can be quickly weakened by using a scalar, while other parts outside the strong edge region are retained, so that the generated image of the strong edge region has more image details.

[0015] In an optional embodiment of the present application, the sharpened image includes a weak edge area image and a strong edge area image, and establishing a guidance map and a noise map based on the reconstructed image and the sharpened image includes: adding the weak edge area image and the reconstructed image to obtain the noise map; adding the strong edge area image and the reconstructed image to obtain the guidance map.

[0016] In an optional embodiment of the present application, denoising the noise map based on the guided graph to generate a denoised image includes: calculating weight parameters of bilateral filtering based on the guided graph; and denoising the noise map based on the weight parameters to generate the denoised image.

[0017] In the above embodiment, the weights of bilateral filtering are calculated through a guided graph to reduce the noise image, thereby enhancing the noise reduction intensity at the strong edges of the image and alleviating the burr feeling at the strong edges of the image after the noise reduction of the edge-preserving noise reduction algorithm. At the same time, sharpening is performed during the noise reduction process to enhance the details of the image.

[0018] In a second aspect, the present application also provides a noise reduction method, which includes: obtaining multiple residual images of an image, each residual image being the difference between the image at different resolutions, and the multiple residual images having a preset layer sequence; superimposing the multiple residual images in sequence multiple times according to the layer sequence to generate a noise-reduced image of the image; wherein at least one superposition in the multiple superpositions includes: weakening the strong edge area in the filtered image corresponding to the residual image of the current layer, and enhancing the weak edge area to generate a sharpened image; obtaining the reconstructed image of the current layer based on the reconstructed image of the previous layer and the sharpened image.

[0019] In the above embodiment, during the superposition process of each layer of residual images of the image to be denoised, the superimposed image is sharpened, which increases the local contrast of the image, enhances the edges and details in the image, and avoids sharpening after subsequent image reconstruction. The implementation process is simpler and more convenient, and the strong edge area is weakened, which can reduce the noise in the strong edge area and reduce the burrs in the image.

[0020] [Corrected on 14.09.2024 according to Rule 91] In an optional embodiment of the present application, the multiple residual images are sequentially superimposed multiple times according to a layer order to generate a denoised image, including: starting with a target residual image among the multiple residual images, performing multiple rounds of iterations according to the layer order until the residual image with the lowest layer order is executed, and stopping the iteration, and the reconstructed image of the last round is the denoised image, and the target residual image refers to the residual image with the third layer order; wherein, the iteration includes: filtering the residual image of the current round to generate a filtered image of the current round; weakening the strong edge area in the filtered image of the current round, and enhancing the weak edge area to generate a sharpened image of the current round; obtaining the reconstructed image of the current round based on the reconstructed image of the previous round and the sharpened image of the current round.

[0021] In the above embodiment, multiple residual images of an image can be quickly reconstructed through an iterative method to obtain a denoised image.

[0022] In an optional embodiment of the present application, the method for generating the reconstructed image of the previous round of the first iteration in the multiple rounds of iteration includes: performing a first filtering on the first residual image to generate a first filtered image, and the first residual image is the residual image with the highest hierarchy; performing a second filtering on the second residual image to generate a second filtered image, and the first filtering is different from the second filtering, and the second residual image is the next layer image of the first residual image; weakening the strong edge area in the second filtered image and enhancing the weak edge area to generate a corresponding second sharpened image; and obtaining the reconstructed image of the previous round of the first iteration in the multiple rounds of iteration based on the first filtered image and the second sharpened image.

[0023] In a third aspect, an embodiment of the present application provides a noise reduction device, which includes: a filtering module for filtering a residual image of an image to generate a filtered image, wherein the residual image is the difference between the image at different resolutions, and the image corresponds to at least two residual images, and there is a preset hierarchical relationship between the at least two residual images; a sharpening module for enhancing the weak edge area in the filtered image and weakening the strong edge area in the filtered image to generate a sharpened image; a generation module for obtaining a target image based on a reconstructed image and the sharpened image, wherein the reconstructed image is the superposition of all residual images in the hierarchical order before the residual image.

[0024] In a fourth aspect, an embodiment of the present application provides an electronic device comprising: a memory and a processor, the processor being connected to the memory; the memory being used to store programs; and the processor being used to call the programs stored in the memory to execute a method as described in any one of the first aspects.

[0025] In a fifth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method as described in any one of the first aspects is executed.

[0026] Other features and advantages of the present application will be described in the following description. The purpose and other advantages of the present application can be realized and obtained through the structures specifically pointed out in the written description and the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] FIG1 is a flow chart of a noise reduction method provided in an embodiment of the present application;

[0028] FIG2 is a flow chart of another noise reduction method provided in an embodiment of the present application;

[0029] FIG3 is a schematic block diagram of a noise reduction device provided in an embodiment of the present application;

[0030] FIG4 is a schematic block diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0031] Hereinafter, embodiments of the present application are described in detail with reference to the accompanying drawings. It should be noted that although the same elements are shown in different drawings, they will be represented by the same reference numerals. In the following description, specific details such as detailed configuration and components are provided to only help a comprehensive understanding of the embodiments of the present application. Therefore, it should be clear to those skilled in the art that various changes and modifications to the embodiments described herein may be made without departing from the scope of the present application. In addition, for the sake of clarity and conciseness, descriptions of well-known functions and configurations have been omitted. The terms described below are defined in consideration of the functions in the present application and may vary according to the user, the user's intention or custom. Therefore, the definition of the terms should be determined based on the content throughout this specification.

[0032] The present application may have various modifications and various embodiments, and in the present application, the embodiments are described in detail below with reference to the accompanying drawings. However, it should be understood that the present application is not limited to the embodiments, but includes all modifications, equivalents and substitutes within the scope of the present application.

[0033] The terms used herein are only used to describe various embodiments of the present application and are not intended to limit the present application. Unless the context clearly indicates otherwise, the singular form is intended to include the plural form. In the present application, it should be understood that the terms "including" or "having" indicate the presence of features, quantities, steps, operations, structural elements, parts or a combination thereof, and do not exclude the presence of one or more other features, quantities, steps, operations, structural elements, parts or a combination thereof, or the possibility of adding one or more other features, quantities, steps, operations, structural elements, parts or a combination thereof.

[0034] Unless defined differently, all terms used herein have the same meaning as understood by those skilled in the art to which this application belongs. Unless explicitly defined in this application, terms (such as those defined in general dictionaries) should be interpreted as having the same meaning as in the context of the relevant art and should not be interpreted as having an idealized or overly formal meaning.

[0035] The electronic device according to one embodiment may be one of various types of electronic devices. The electronic device may include, for example, a portable communication device (e.g., a smartphone), a computer, a portable multimedia device, a portable medical device, a camera, a wearable device, or a household appliance. According to one embodiment of the disclosure, the electronic device is not limited to the above examples.

[0036] The term used in this application is not intended to limit the application, but is intended to include the various changes, equivalents or substitutes of corresponding embodiments. About the description of the accompanying drawings, similar reference numerals can be used to represent similar elements or related elements. Unless the relevant context clearly indicates otherwise, the singular form of the noun corresponding to the item can include one or more things. As used herein, each of the phrases such as "A or B", "at least one of A and B", "at least one of A or B", "A, B or C", "at least one of A, B and C" and "at least one of A, B or C" can include all possible combinations of the item enumerated together in the corresponding one in the phrase. As used herein, terms such as "the 1st", "the 2nd", "first" and "second" can be used to distinguish corresponding components from another component, and are not intended to limit components in other aspects (for example, importance or order). It is intended that if an element (e.g., a first element) is referred to as being “coupled with,” “coupled to,” “connected to,” or “connected to” another element (e.g., a second element), with or without the term “operably” or “communicatively,” it indicates that the element may be coupled with the other element directly (e.g., wired), wirelessly, or via a third element.

[0037] As used herein, the term "module" may include a unit implemented in hardware, software, or firmware, and may be used interchangeably with other terms such as "logic," "logic block," "component," and "circuit." A module may be a single integrated component or its smallest unit or component adapted to perform one or more functions. For example, according to one embodiment, a module may be implemented in the form of an application-specific integrated circuit (ASIC).

[0038] As shown in FIG1 , FIG1 is a flowchart of a noise reduction method shown in an embodiment of the present application. As shown in FIG1 , the method includes steps S110 to S130:

[0039] S110 , filtering a residual image of an image to generate a filtered image, wherein the residual image is a difference between the image at different resolutions, the image corresponds to at least two residual images, and a preset hierarchical relationship exists between the at least two residual images.

[0040] A residual image of the image is obtained. The residual image is any residual image in a residual image set corresponding to the image. Each residual image in the residual image set is a difference between the image at different resolutions, and the residual images in the residual image set have a predetermined hierarchical relationship. For ease of distinction, in this embodiment, the obtained residual image is referred to as the first residual image.

[0041] For example, the multiple residual images corresponding to the image to be denoised can be generated using a Laplacian pyramid.

[0042] Methods for generating residual image sets using the Laplacian pyramid include:

[0043] 1. Construct a Gaussian pyramid. Assume that the original resolution of the image to be denoised is 64×64. The image to be denoised serves as the bottom layer of the Gaussian pyramid (G-Level 0). This image is filtered, for example, by applying a Gaussian filter. Gaussian filtering is a method for blurring an image, which reduces high-frequency information and is primarily used for denoising and smoothing. For example, convolving the denoised image with a Gaussian kernel can achieve Gaussian filtering. Next, downsample the Gaussian-filtered image. For example, nearest neighbor interpolation can be used: every other pixel in the image is used as the pixel of the new image, reducing the original image to a 32×32 image. This new image constitutes the upper layer of the Gaussian pyramid, the second-level image (G-Level 1). Filtering and downsampling the second-level image (G-Level 1) yields a new image (G-Level 2). Repeat this process until the image to be denoised reaches the specified size, or until the preset number of layers is reached, at which point execution stops. After stopping execution, a Gaussian pyramid of the image to be denoised can be obtained, which contains images of multiple resolutions, from high-resolution images to images of increasingly lower resolutions from bottom to top. In the embodiment of the present application, the filtering method is not limited to Gaussian filtering, and it can also be other filtering methods. Similarly, the method for downsampling the Gaussian pyramid can also be bilinear downsampling, area sampling, maximum pooling, etc., and the present application does not limit the downsampling method.

[0044] [Corrected 14.09.2024 in accordance with Rule 91] 2. Construct a Laplacian pyramid based on the obtained Gaussian pyramid. Each level of the Laplacian pyramid is constructed by subtracting the upsampled, Gaussian-blurred image of the previous level from the first level of the Gaussian pyramid. The Laplacian pyramid construction method includes: starting with the image at the second level of the Gaussian pyramid (G-Level 1), upsampling the second level image to obtain a 64×64 image, Gaussian-blurring the upsampled image to obtain a blurred image, subtracting this blurred image from the image at the first level of the Gaussian pyramid (G-Level 0), and obtaining the bottommost image of the Laplacian pyramid (L-Level 0), which is a residual image. Upsampling and Gaussian-blurring the image at the third level of the Gaussian pyramid (G-Level 2) to generate a blurred image, and subtracting this blurred image from the image at the second level of the Gaussian pyramid to obtain the image at the second level of the Laplacian pyramid (L-Level 1). This process is repeated until the last level of the Gaussian pyramid is reached, at which point the Laplacian pyramid of the image is obtained.

[0045] Each image in the Laplacian pyramid is a residual image, the difference between images to be denoised at different resolutions. All residual images in the Laplacian pyramid correspond to the image to be denoised, and all residual images in the Laplacian pyramid have a preset layer order. Image reconstruction can be achieved using the Laplacian pyramid: starting with the smallest-scale Laplacian pyramid image, it is upsampled and added to the Laplacian pyramid image in the next layer. The superposition result is then upsampled and added to the Laplacian pyramid image in the next layer. This superposition is repeated along the descending order of the layers until the bottom layer of the Laplacian pyramid is reached. By superimposing all residual images, the image to be denoised can be reconstructed. During the reconstruction process, denoising is also performed on the image, for example, by applying a Gaussian filter before upsampling.

[0046] For example, a Laplacian pyramid corresponding to the image to be denoised is obtained, and any layer of image in the Laplacian pyramid is selected as the first residual image. For example, the third layer of image in the Laplacian pyramid is selected as the first residual image.

[0047] The first residual image is filtered to generate a corresponding first filtered image. The filtering may be bilateral filtering, Gaussian filtering, etc., and this application does not limit the filtering method. In one embodiment of the present application, Gaussian filtering is used for filtering: a 3*3 Gaussian kernel with a bandwidth of σ1 is used to filter the first residual image. The purpose of convolution is noise reduction, which can prevent strong and weak edges in the residual image from being affected by noise.

[0048] S120 , enhancing the weak edge region in the filtered image and weakening the strong edge region in the filtered image to generate a sharpened image.

[0049] In image processing, an edge is a region of image brightness that exhibits significant changes, typically the boundaries of objects or features. Edge strength generally refers to the magnitude of the change in pixel values ​​in that region of the image. Strong edges correspond to areas of large brightness changes and are typically clear, prominent boundaries or contours, representing the primary outlines of objects or features in the image. Weak edges correspond to areas of smaller brightness changes and may be blurred or unclear boundaries.

[0050] For example, a strong edge region or a weak edge region may be distinguished by calculating the gradient of pixel values, for example, using a Sobel operator or a simple difference.

[0051] The weak edge regions in the first filtered image are enhanced. For example, a high-pass filter or a Laplacian filter can be used to enhance high-frequency information in the image, thereby enhancing the weak edges. Alternatively, contrast adjustment or non-local methods can be used for enhancement. The specific enhancement methods are not limited in the present embodiment.

[0052] Moreover, the strong edge regions in the first filtered image are weakened. For example, the bilateral filtering method is used to smooth the strong edges, thereby achieving the weakening of the strong edges. For another example, the present application can also use methods such as Gaussian blur or waveform compression to weaken the strong edge regions. The embodiments of the present application do not limit the specific means for weakening.

[0053] By enhancing the weak edge regions, the high-frequency part of the image is enhanced, thereby increasing the local contrast of the image, enhancing the edges and details in the image, and making the image look clearer and sharper. At the same time, weakening the strong edge regions can reduce the noise in the strong edge regions and weaken the burrs in the image.

[0054] In an embodiment of the present application, the method further includes: assigning values to the filtered image based on a preset threshold interval to obtain the weak edge regions; assigning values to the filtered image based on a preset threshold to obtain the strong edge regions.

[0055] The strong edge regions can be obtained through thresholds. The methods for obtaining the strong edge regions and the weak edge regions include: presetting edge detection thresholds: th1, th2. For example, it is determined that the flat region is: pixel value < th1, the weak edge region is determined as: th1 <= pixel value < th2, and the strong edge region is determined as: pixel value >= th2.

[0056] Further, the result after filtering the first residual image is denoted as ResIntensity1 after taking the absolute value. For example, by convolving the first residual image with a predefined kernel function (kernel) and then taking the absolute value, ResIntensity1 is obtained, and this value represents the intensity of the residual layer. Taking the absolute value can cancel the positive and negative signs of the values in the residual map. A single value cannot indicate whether the current pixel is in the strong edge region, the weak edge region, or the flat region.

[0057] If a 3*3 Gaussian kernel with a bandwidth of σ1 is used to filter Pyr[3], the filtering can be expressed as Equation 1:

[0058] Where, kernel represents the Gaussian kernel, and pyr[3] refers to the above-mentioned first residual image. In this embodiment, the third-layer image in the Laplacian pyramid is taken, so it is represented as pyr[3].

[0059] According to ResIntensity1 and the edge detection threshold, it can be distinguished whether each pixel in the current residual layer belongs to the flat region, the weak edge region, or the strong edge region.

[0060] For the acquisition of weak edge regions, a method of assigning values to ResIntensity1 can be adopted. For example, all pixels with th1 <= pixel value < th2 are assigned a value of 0. Other pixels are assigned a value of 1, which can be expressed as Formula 2:

[0061] where (x, y) represents pixel coordinates.

[0062] By assigning values to ResIntensity1, an image of weak edge regions can be generated.

[0063] For the acquisition of strong edge regions, a method of assigning values to ResIntensity1 can also be adopted. For example, all pixels with pixel value >= th2 are assigned a value of 1. Other pixels are assigned a value of 0, which can be expressed as Formula 3:

[0064] where (x, y) represents pixel coordinates.

[0065] By assigning values to ResIntensity1, an image of strong edge regions is generated.

[0066] In the above embodiments, by assigning values through preset thresholds, the strong edge regions and weak edge regions of the image can be obtained quickly and simply.

[0067] In an embodiment of the present application, the sharpened image includes an image of weak edge regions, and the enhancement of the weak edge regions in the filtered image includes: using a preset first scalar to enhance the weak edge regions and retaining other parts outside the weak edge regions in the first filtered image to generate the image of weak edge regions.

[0068] To enhance the residual layer of the weak edge regions, the following Formula 4 can be used: NewRes1 = Gain1 * Pyr[3] * WeakEdge + Pyr[3] * (1 - WeakEdge) Formula 4

[0069] where Gain1 is a preset scalar, and Gain1 > 1. Pyr[3] refers to the above-mentioned first residual image. In this embodiment, the third-layer image in the Laplacian pyramid is taken, so it is represented as Pyr[3], and WeakEdge represents the weak edge regions. For example, it can be the image of weak edge regions obtained by assignment.

[0070] In formula 4, the scalar (i.e., Gain1) is multiplied with each pixel value in the first residual image Pyr[3] and the weak edge area WeakEdge to obtain an image of the same size as the original image, in which the weak edge area in the first residual image is linearly enlarged by the scalar. Further, Pyr[3]*(1-WeakEdge) is added to the linearly enlarged image to finally generate a weak edge area image. By adding, the other parts outside the weak edge area are retained in the final output image, and only the weak edge area is enlarged. It should be noted that when the weak edge area is a separate weak edge area image, the enhanced weak edge area image is part of the sharpened image.

[0071] In the above embodiment, the weak edge region can be quickly enhanced by using a scalar, while other parts outside the weak edge region are retained, so that the generated image of the weak edge region has more image details.

[0072] In one embodiment of the present application, the sharpened image includes a strong edge area image, and the weakening of the strong edge area in the filtered image includes: using a preset second scalar to weaken the strong edge area, and retaining other parts outside the strong edge area in the filtered image to generate the strong edge area image.

[0073] To enhance the residual layer in the strong edge area, the following formula 5 can be used: NewRes2 = Gain2*Pyr[3]*StrongEdge+Pyr[3]*(1-StrongEdge) Formula 5

[0074] Wherein, Gain2 is a preset scalar, 0<Gain2<1. Pyr[3] refers to the first residual image mentioned above. In this embodiment, the third layer image in the Laplacian pyramid is taken, so it is expressed as Pyr[3]. StrongEdge represents a strong edge region. For example, it can be a strong edge region image obtained by assignment.

[0075] In formula 5, the scalar (i.e., Gain2) is multiplied with each pixel value in the first residual image Pyr[3] and the strong edge region StrongEdge, resulting in an image of the same size as the original image, in which the strong edge region in the first residual image is linearly reduced by the scalar. Further, Pyr[3]*(1-StrongEdge) is added to the linearly reduced image to finally generate a strong edge region image. By adding, the other parts outside the strong edge region are retained in the final output image, and only the strong edge region is reduced. It should be noted that when the strong edge region is an independent strong edge region image, the weakened strong edge region image is part of the sharpened image.

[0076] In the above embodiment, the strong edge region can be quickly weakened by using a scalar, while other parts outside the strong edge region are retained, so that the generated image of the strong edge region has more image details.

[0077] S130 , obtaining a target image according to the reconstructed image and the sharpened image, wherein the reconstructed image is a superposition of all residual images whose layer sequence is before the residual image.

[0078] Obtain a reconstructed image. The reconstructed image refers to the reconstruction result of the residual image whose layer sequence precedes the layer sequence of the first residual image. For example, assuming that the Laplacian pyramid has six layers and the first residual image is the third layer image in the Laplacian pyramid, the reconstructed image is the result of reconstructing the sixth layer image, the fifth layer image, and the fourth layer image in the Laplacian pyramid. The reconstruction may include image sharpening, stacking, and noise reduction. For example, the noise reduction method of this embodiment is used to perform image stacking on the residual images reconstructed into adjacent layers.

[0079] Based on the reconstructed image and the sharpened image, a target image can be obtained by reconstructing the image to be denoised. For example, the target image can be obtained by superimposing the reconstructed image and the sharpened image. The target image can be an intermediate image in the denoising process. For example, the target image needs to be further superimposed with the image below the third layer in the Laplacian pyramid to obtain the denoised image to be denoised. Alternatively, the target image can be the denoised image to be denoised. For example, when the Laplacian pyramid has only three layers, the resulting image is the denoised image to be denoised.

[0080] During the reconstruction process, the superimposed images are sharpened and the weak edge areas of the superimposed images are enhanced, thereby enhancing the high-frequency portion of the image. This increases the local contrast of the image, enhances the edges and details in the image, and makes the image appear clearer and sharper. At the same time, the strong edge areas are weakened to reduce noise in the strong edge areas and reduce image glitches.

[0081] In the above embodiment, sharpening is performed during the reconstruction process of the image to be denoised, which increases the local contrast of the image and enhances the edges and details in the image. There is no need to sharpen the image after reconstruction, which makes the implementation process simpler and more convenient. In addition, the strong edge area is weakened, which can reduce the noise in the strong edge area and reduce the burrs in the image.

[0082] In one embodiment of the present application, obtaining a target image based on the reconstructed image and the sharpened image includes: establishing a guidance map and a noise map based on the reconstructed image and the sharpened image; denoising the noise map based on the guidance map to obtain a denoised image; and upsampling the denoised image to obtain the target image.

[0083] [Corrected 14.09.2024 in accordance with Rule 91] A guidance map and a noise map are respectively established based on the reconstructed image and the sharpened image. The guidance map is an image used to reduce noise in the noise map, and the noise map is a superposition of part or all of the reconstructed image and the sharpened image. Noise reduction is performed on the noise map based on the guidance map to obtain a reduced-noise image. The reduced-noise image is then upsampled to obtain the target image.

[0084] Furthermore, the sharpened image includes a weak edge area image and a strong edge area image, and establishing a guidance map and a noise map based on the reconstructed image and the sharpened image includes: adding the weak edge area image and the reconstructed image to obtain the noise map; adding the strong edge area image and the reconstructed image to obtain the guidance map.

[0085] The sharpened image includes the image in the weak edge area and the image in the strong edge area. The noise image is recorded as NoisyImage and the guide image is recorded as GuideImage. The establishment process of the two can be expressed as Formula 6: NoisyImage = Up1 + NewRes1 Formula 6 GuideImage = Up1 + NewRes2 Formula 6

[0086] Among them, NewRes1 represents the image in the weak edge area, NewRes2 represents the image in the strong edge area, and Gp1 represents the reconstructed image, which usually needs to be upsampled. For example, when the above-mentioned first residual image is the third layer image in the Laplace pyramid, the reconstructed image is the upsampling of the reconstruction result of the sixth layer image, the fifth layer image, and the fourth layer image in the Laplace pyramid. The amount of upsampling is determined by the size of the residual image, that is, the upsampled image is the same size as the residual image, which facilitates the superposition of the reconstructed image and the sharpened image.

[0087] In the above embodiment, the guidance map and the noise map can be established quickly and easily. The use of the guidance map can give more weight to the strong edge area, thereby enhancing the noise reduction strength at the strong edges in the image. The noise map is the result of superimposing the residual image after sharpening, which increases the local contrast of the image, enhances the edges and details in the image, and can make the generated image clearer and sharper.

[0088] In one embodiment of the present application, denoising the noise map based on the guided graph to generate a denoised image includes: calculating weight parameters of bilateral filtering based on the guided graph; and denoising the noise map based on the weight parameters to generate the denoised image.

[0089] The weights of bilateral filtering are calculated using the guided graph to reduce the noise image and generate a reduced-noise image. Bilateral filtering is a process that determines the weighted values ​​of points in the neighborhood of the pixel point (x, y) during filtering by calculating the pixel difference Img(x, y)-Img(x+k1, y+k2) between the pixel value of the current point (x, y) and the pixel values ​​of other points in its neighborhood (x+k1, y+k2) and the position difference of the pixels.

[0090] The noise reduction result is recorded as NR2, and the noise reduction process can be expressed as Formula 7:

[0091] [Corrected 14.09.2024 according to Rule 91] Where k1 represents the offset value of the pixel's row coordinates, and the range of k1 is [-r, r]. k2 represents the offset value of the pixel's column coordinates, and the range of k2 is [-r, r]. r represents the radius of the value. x+j1,y+k2 represents the points within the radius r of the current point (x, y). NoisyImage represents a noise image, e is an exponential function, and σ1 and σ2 both represent the bandwidth of the Gaussian function, which is used to control the strength of the noise reduction. σ1 is the bandwidth of the Gaussian function related to the pixel value, and σ2 is the bandwidth of the Gaussian function related to the distance between the pixel coordinates.

[0092] NR2(x,y) represents the denoised image, which is the weight calculated by bilateral filtering. The weighted average of the points in the neighborhood of (x,y) centered at the pixel point (x,y) is used as the denoising result of the point (x,y).

[0093] In the above embodiment, the weights of bilateral filtering are calculated through a guided graph to reduce the noise image, thereby enhancing the noise reduction intensity at the strong edges of the image and alleviating the burr feeling at the strong edges of the image after the noise reduction of the edge-preserving noise reduction algorithm. At the same time, sharpening is performed while reducing the noise, thereby enhancing the details of the image.

[0094] The embodiment shown in FIG1 is applicable to processing the residual image of one adjacent layer during image reconstruction. If other methods are used for other residual images of the image, the noise reduction image finally reconstructed may not achieve the optimal effect. Based on this, the present application also provides a second embodiment shown in FIG2, in which the noise reduction method shown in FIG1 is applied to the residual image of each adjacent layer, so that the reconstructed image has a better effect.

[0095] As shown in FIG2 , FIG2 is a flowchart of another noise reduction method shown in an embodiment of the present application. As shown in FIG2 , the method includes steps S210 to S220:

[0096] S210, acquiring a plurality of residual images of an image, each residual image being a difference value of the image at different resolutions, and the plurality of residual images having a preset layer sequence;

[0097] [Corrected 14.09.2024 in accordance with Rule 91] Obtain all residual images corresponding to the image to be denoised, where each residual image is a difference between the image at different resolutions, and the residual images have a preset hierarchical relationship. For example, obtain a Laplacian pyramid of the image to be denoised, where the Laplacian pyramid is composed of multiple residual images, and the residual images in the Laplacian pyramid have a preset hierarchical relationship.

[0098] S220, the multiple residual images are sequentially superimposed multiple times according to the layer order to generate a denoised image of the image; wherein, at least one superposition in the multiple superpositions includes: weakening the strong edge area in the filtered image corresponding to the residual image of the current layer, and enhancing the weak edge area to generate a sharpened image; and obtaining the reconstructed image of the current layer based on the reconstructed image of the previous layer and the sharpened image.

[0099] All residual images corresponding to the image to be denoised are superimposed in sequence according to the layer order to generate a denoised image. For example, if the residual image has three layers, the third-level image is superimposed with the second-level image to obtain a superposition result, and the superposition result is superimposed with the first-level image to generate the denoised image to be denoised.

[0100] Among them, in the process of superimposing images of some adjacent layers: obtain the filtered image of the current layer sequence, and this filtered image is obtained by filtering based on the residual image of the current layer sequence. For example, there are six residual images in total, and the superposition starts from the top layer. When it reaches the third layer, obtain the residual image of the current layer, that is, the third layer, and filter the image to generate the filtered image of the current layer. Weaken the strong edge area in the filtered image of the current layer, and enhance the weak edge area to generate a sharpened image of the current layer. Obtain the reconstructed image of the layer above the current layer. The reconstructed image is the superposition of all images above the current layer. For example, when the current layer is the residual image of the third layer, obtain the reconstructed image of the previous layer, that is, the fourth layer. The reconstructed image of the fourth layer is the image generated by processing and superimposing the residual images of the sixth, fifth, and fourth layers. Reconstruct the reconstructed image of the current layer based on the reconstructed image of the previous layer and the sharpened image of the current layer.

[0101] In the above embodiment, during the superposition process of each layer of residual images of the image to be denoised, the superimposed image is sharpened, which increases the local contrast of the image and enhances the edges and details in the image. There is no need to perform sharpening after subsequent image reconstruction, and the implementation process is simpler and more convenient. In addition, the strong edge area is weakened, which can reduce the noise in the strong edge area and reduce the burrs in the image.

[0102] [Corrected on 14.09.2024 according to Rule 91] In one embodiment of the present application, the multiple residual images are superimposed multiple times in sequence according to the layer order to generate a denoised image, including: starting with a target residual image among the multiple residual images, performing multiple rounds of iterations according to the layer order until the residual image with the lowest layer order is executed, and stopping the iteration, and the reconstructed image of the last round is the denoised image, and the target residual image refers to the residual image with the third layer order; wherein, the iteration includes: filtering the residual image of the current round to generate a filtered image of the current round; weakening the strong edge area in the filtered image of the current round, and enhancing the weak edge area to generate a sharpened image of the current round; obtaining the reconstructed image of the current round based on the reconstructed image of the previous round and the sharpened image of the current round.

[0103] Residual image processing and superposition can be performed in an iterative manner. The iterative methods include:

[0104] [Corrected 14.09.2024 according to Rule 91] Starting from a target residual image among multiple residual images, multiple rounds of iterations are performed in order from large to small in layer order, wherein the target residual image refers to the image from which iteration is started, which is usually the third residual layer in the layer order of the residual image. In other embodiments of the present application, it can also be other layers of the residual image. The residual image in the third layer order is used as the target residual image, and iteration is performed starting from the target residual image until the iteration is completed for the residual image with the lowest layer order, and the iteration is stopped. At this point, the reconstructed image generated during the last round of iteration is the denoised image. For example, the residual image is in the form of a Laplacian pyramid. The Laplacian pyramid has a total of six layers. From the top to the bottom, the layer order numbers of the residual images are 5 to 0, and the iteration will be performed starting from the third residual image in the layer order (i.e., the residual image numbered 3) until the residual image in the bottom layer (i.e., the residual image numbered 0) is executed.

[0105] The iteration includes: obtaining the residual image of the current round of iteration. For example, assuming that the current round is the first round of iteration, the residual image layer sequence of the current round is the third residual image, obtaining the residual image and filtering the image to generate a filtered image of the current round of iteration.

[0106] The strong edge areas in the current round of filtered images are weakened, and the weak edge areas are enhanced to generate the current round of sharpened images.

[0107] Obtain the reconstructed image from the previous iteration. Each reconstructed image is the result of processing and superimposing all residual images prior to the current iteration's residual image. If the current iteration's residual image is the third-order residual image, the previous iteration's reconstructed image is the result of processing and superimposing the fourth-order and fifth-order residual images. Furthermore, upsample the previous iteration's reconstructed image to the same size as the current iteration's filtered image.

[0108] The reconstructed image of the previous round is processed and superimposed with the sharpened image of the current round to generate the reconstructed image of the current round.

[0109] In the above embodiment, multiple residual images of an image can be quickly reconstructed through an iterative method to obtain a denoised image.

[0110] In one embodiment of the present application, a method for generating a reconstructed image of the previous round of the first iteration in the multiple rounds of iteration includes: performing a first filtering on the first residual image to generate a first filtered image, the first residual image being the residual image with the highest hierarchy; performing a second filtering on the second residual image to generate a second filtered image, the first filtering is different from the second filtering, and the second residual image is the next layer image of the first residual image; weakening the strong edge area and enhancing the weak edge area in the second filtered image to generate a corresponding second sharpened image; and obtaining the reconstructed image of the previous round of the first iteration in the multiple rounds of iteration based on the first filtered image and the second sharpened image.

[0111] If no image is reconstructed in the first round of iteration, the method for generating the reconstructed image of the previous round used in the first round of iteration includes:

[0112] Obtain the residual image with the highest order among multiple residual images, such as the six-layer Laplacian pyramid used in the above iteration, and obtain the residual image of the sixth layer. Perform a first filter on the residual image of this layer to obtain a first filtered image. The specific means of the first filter are not limited in this application. The first filter can be a bilateral filter, a Gaussian filter, etc. Obtain the next layer of the residual image with the highest order, for example, obtain the next layer of the image of the sixth layer, that is, the fifth layer, and perform a second filter on the image to generate a second filtered image. The second filter can be a Gaussian filter or a bilateral filter. When the first filter and the second filter are inconsistent, the present application can achieve better implementation results. For example, in one embodiment of the present application, the first filter is a bilateral filter, and the second filter is a Gaussian filter.

[0113] The strong edge region in the second filtered image is weakened, and the weak edge region in the second filtered image is enhanced to generate a corresponding second sharpened image.

[0114] Reconstruction is performed based on the second sharpened image and the first filtered image to generate the reconstructed image from the previous iteration, used in the first iteration. Furthermore, the first filtered image needs to be upsampled before reconstruction. The amount of upsampling is determined by the size of the second residual image. That is, the image generated by upsampling the first filtered image should be the same size as the second filtered image. For example, when the second filtered image is twice the size of the first filtered image, bilinear interpolation is used to upsample the first filtered image by a factor of two to generate the first enlarged image.

[0115] [Corrected 14.09.2024 according to Rule 91] The above-mentioned reconstruction includes image sharpening, superposition, noise reduction, etc. For example, the absolute value of the second superimposed image is taken to obtain a second absolute value, and the second absolute value is assigned according to a preset edge detection threshold to generate a strong edge layer and a weak edge layer, respectively. The weak edge layer is enhanced to generate a weak edge image, and the strong edge layer is weakened to generate a strong edge image. The weak edge image is added to the first amplified image to obtain a noise map, and the strong edge image is added to the first amplified image to obtain a guide map. The guide map is used to calculate the weights of bilateral filtering, and the noise map is bilaterally filtered based on the obtained weights to generate the reconstructed image of the previous round used in the first round of iteration.

[0116] As shown in FIG3 , FIG3 shows a structural block diagram of a noise reduction module 300 provided in an embodiment of the present application. The noise reduction module 300 includes: a filtering module 310 , a sharpening module 320 and a generating module 330 .

[0117] The filtering module 310 is used to filter the residual image of the image to generate a filtered image, where the residual image is the difference between the image at different resolutions, the image corresponds to at least two residual images, and there is a preset hierarchical relationship between the at least two residual images.

[0118] The sharpening module 320 is configured to enhance the weak edge regions in the filtered image and weaken the strong edge regions in the filtered image to generate a sharpened image.

[0119] The generating module 330 is configured to obtain a target image according to the first filtered image and the sharpened image.

[0120] In an optional embodiment of the present application, the generating module 330 includes:

[0121] An image generation submodule, configured to establish a guidance map and a noise map based on the reconstructed image and the sharpened image;

[0122] a denoising submodule, configured to denoise the noise image according to the guide image to obtain a denoised image;

[0123] The upsampling submodule is used to upsample the denoised image to obtain the target image.

[0124] In an optional embodiment of the present application, the noise reduction module 300 further includes:

[0125] A first assignment module, configured to assign a value to the filtered image based on a preset threshold interval to obtain the weak edge region;

[0126] The second assignment module is used to assign a value to the filtered image based on a preset threshold to obtain the strong edge area.

[0127] In an optional embodiment of the present application, the sharpened image includes a weak edge area image, and the sharpening module 320 is specifically used to enhance the weak edge area using a preset first scalar, and retain other parts outside the weak edge area in the filtered image to generate the weak edge area image.

[0128] In an optional embodiment of the present application, the sharpened image includes a strong edge area image, and the sharpening module 320 is specifically used to use a preset second scalar to weaken the strong edge area and retain other parts outside the strong edge area in the filtered image to generate the strong edge area image.

[0129] In an optional embodiment of the present application, the sharpened image includes a weak edge area image and a strong edge area image, and the image generation submodule is specifically used to add the weak edge area image and the reconstructed image to obtain the noise map; and add the strong edge area image and the reconstructed image to obtain the guidance map.

[0130] In an optional embodiment of the present application, the denoising submodule is specifically configured to calculate weight parameters of bilateral filtering based on the guided graph; and denoise the noise graph according to the weight parameters to generate the denoised image.

[0131] As shown in Figure 4, Figure 4 shows a structural block diagram of an electronic device 400 provided in an embodiment of the present application. The electronic device 400 includes: a processor 410 and a memory 420;

[0132] The processor 410, the memory 420 and each component are electrically connected to each other directly or indirectly to realize data transmission or interaction. For example, these components can be electrically connected to each other through one or more communication buses or signal lines. The memory 420 is used to store computer programs, such as the software function module shown in Figure 3, i.e., the noise reduction module 300. Among them, the noise reduction module 300 includes at least one software function module that can be stored in the memory 420 in the form of software or firmware or solidified in the operating system (OS) of the electronic device 400. The processor 410 is used to execute the executable module stored in the memory 420, such as the software function module or computer program included in the noise reduction module 300.

[0133] Among them, the memory 420 can be, but is not limited to, random access memory (RAM), read only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), etc.

[0134] The processor 410 may be an integrated circuit chip with signal processing capabilities. The above-mentioned processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), a microprocessor, etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The various methods, steps and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. Alternatively, the processor 410 may also be any conventional processor, etc.

[0135] An embodiment of the present application further provides a non-volatile computer-readable storage medium (hereinafter referred to as storage medium), on which a computer program is stored. When the computer program is run by a computer such as the electronic device 400 described above, the noise reduction method shown above is executed.

[0136] It should be noted that the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same and similar parts between the various embodiments can be referenced to each other.

[0137] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architectures, functions and operations of the devices, methods and computer program products according to the multiple embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of the code, and the module, program segment or a part of the code contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or can be implemented using a combination of dedicated hardware and computer instructions.

[0138] In addition, the functional modules in each embodiment of the present application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0139] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a computer-readable storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, a laptop, a server, or an electronic device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned computer-readable storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0140] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A noise reduction method, characterized in that: The method comprises: Filtering a residual image of an image to generate a filtered image, wherein the residual image is a difference between the image at different resolutions, the image corresponds to at least two residual images, and a preset hierarchical relationship exists between the at least two residual images; enhancing the weak edge region in the filtered image and weakening the strong edge region in the filtered image to generate a sharpened image; A target image is obtained according to the reconstructed image and the sharpened image, wherein the reconstructed image is a superposition of all residual images whose layer sequence is before the residual image.

2. The method according to claim 1, characterized in that Obtaining a target image according to the reconstructed image and the sharpened image includes: Establishing a guidance map and a noise map based on the reconstructed image and the sharpened image; Denoising the noise image according to the guidance graph to obtain a denoised image; The denoised image is upsampled to obtain the target image.

3. The method according to claim 1, characterized in that The method further comprises: Assigning a value to the filtered image based on a preset threshold interval to obtain the weak edge area; The filtered image is assigned a value based on a preset threshold to obtain the strong edge region.

4. The method according to claim 1, wherein The sharpened image includes a weak edge region image, and the enhancing the weak edge region in the filtered image includes: The weak edge region is enhanced using a preset first scalar, and other parts outside the weak edge region in the filtered image are retained to generate the weak edge region image.

5. The method according to claim 1, wherein The sharpened image includes an image with a strong edge region, and the weakening of the strong edge region in the filtered image includes: The strong edge region is weakened using a preset second scalar, and other parts outside the strong edge region in the filtered image are retained to generate the strong edge region image.

6. The method according to claim 2, characterized in that The sharpened image includes a weak edge region image and a strong edge region image, and establishing a guide map and a noise map based on the reconstructed image and the sharpened image includes: Adding the weak edge region image and the reconstructed image to obtain the noise map; The strong edge region image and the reconstructed image are added to obtain the guidance map.

7. The method according to claim 2, characterized in that The step of reducing noise on the noise map according to the guide map to generate a reduced noise image includes: Calculating weight parameters of bilateral filtering based on the guided graph; The noise map is denoised according to the weight parameters to generate the denoised image.

8. A noise reduction method, characterized in that: The method comprises: Acquire multiple residual images of an image, each residual image being a difference value of the image at different resolutions, and the multiple residual images having a preset layer sequence; Superimposing the plurality of residual images in sequence for multiple times according to a layer order to generate a denoised image of the image; Among them, at least one superposition among the multiple superpositions includes: weakening the strong edge area in the filtered image corresponding to the residual image of the current layer, and enhancing the weak edge area to generate a sharpened image; obtaining the reconstructed image of the current layer based on the reconstructed image of the previous layer and the sharpened image.

9. [Corrected 14.09.2024 according to Rule 91] The method according to claim 8, characterized in that The step of sequentially stacking the plurality of residual images multiple times according to a layer order to generate a noise-reduced image includes: Starting with a target residual image among the plurality of residual images, performing multiple rounds of iterations according to the layer order until the residual image with the lowest layer order is reached, and then stopping the iterations. The reconstructed image of the last round is the denoised image, and the target residual image is the residual image with the third layer order; The iteration includes: filtering the residual image of the current round to generate the filtered image of the current round; weakening the strong edge area and enhancing the weak edge area in the filtered image of the current round to generate the sharpened image of the current round; obtaining the reconstructed image of the current round based on the reconstructed image of the previous round and the sharpened image of the current round.

10. [Corrected 14.09.2024 according to Rule 91] The method according to claim 9, characterized in that The method for generating the reconstructed image of the previous round of the first iteration in the multiple rounds of iterations includes: Performing a first filtering on the first residual image to generate a first filtered image, wherein the first residual image is the residual image with the highest layer order; performing a second filtering on the second residual image to generate a second filtered image, wherein the first filtering is different from the second filtering, and the second residual image is an image of a next layer of the first residual image; weakening the strong edge region and enhancing the weak edge region in the second filtered image to generate a corresponding second sharpened image; A reconstructed image of a previous round of a first iteration in the multiple rounds of iterations is obtained according to the first filtered image and the second sharpened image.

11. A noise reduction device, characterized in that: The device comprises: a filtering module configured to filter a residual image of an image to generate a filtered image, wherein the residual image is a difference between the image at different resolutions, the image corresponds to at least two residual images, and a preset hierarchical relationship exists between the at least two residual images; a sharpening module, configured to enhance weak edge regions in the filtered image and weaken strong edge regions in the filtered image to generate a sharpened image; A generating module is used to obtain a target image according to the first filtered image and the sharpened image.

12. An electronic device, characterized in that: include: processor and memory; The processor and the memory communicate with each other via a bus; The memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute the noise reduction method according to any one of claims 1 to 10.

13. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer program instructions, and when the computer program instructions are read and executed by a computer, the noise reduction method according to any one of claims 1 to 10 is executed.

Images