Image denoising method for motion region

Abstract

The application provides an image denoising method for a motion region. For the motion region, the pixel gradient of the pixel in the motion region of a target image and the pixel gradient of the pixel in the motion region of a reference image are used to perform time domain denoising on the motion region of the target image to obtain a time domain denoising result. The time domain denoising result is fused with the corresponding spatial domain denoising result of the motion region in the spatial domain denoising result of the target image. The time domain denoising is based on the information of the pixel gradient, the edge and texture structure represented by the pixel gradient can be perceived in the time domain denoising process, the effect of the time domain denoising is improved, and the corresponding spatial domain denoising result can be fused on the basis of the time domain denoising result. The details of the motion region can be effectively preserved while denoising, the clarity of the image is improved, and the image denoising effect is further improved.

Description

Technical Field

[0001] This application relates to the field of image processing technology, and in particular to an image noise reduction method for moving regions. Background Technology

[0002] In the field of image processing technology, noise reduction processing can be performed on images to improve their visual quality.

[0003] When dealing with moving regions in an image, related technologies often use spatial domain noise reduction to denoise the moving regions. This method only uses a single frame of pixels for smoothing, which usually causes image blurring and results in poor image noise reduction. Summary of the Invention

[0004] This application provides an image denoising method for moving regions, which can improve the image denoising effect.

[0005] According to a first aspect of the embodiments of this application, an image denoising method for moving regions is provided, comprising: Acquire the target image and a reference image, where the reference image is the previous frame of the target image; Spatial domain denoising is performed on the target image to obtain the spatial domain denoising result; Based on the pixel difference between pixels at the same position in the target image and the reference image, the motion region in the target image and the reference image is determined; Based on the pixel gradients of pixels in the motion region of the target image and the pixel gradients of pixels in the motion region of the reference image, temporal denoising is performed on the motion region of the target image to obtain the temporal denoising result. The pixel gradient represents the change of pixel value in the spatial dimension. The spatial denoising results corresponding to the moving regions in the temporal and spatial denoising results are fused to obtain the denoised output result of the moving regions of the target image.

[0006] In some embodiments of this application, temporal denoising is performed on the motion region of the target image based on the pixel gradient of pixels in the motion region of the target image and the pixel gradient of pixels in the motion region of the reference image to obtain a temporal denoising result, including: Based on the pixel gradients of each pixel in the motion region of the target image and the pixel gradients of each pixel in the motion region of the reference image, the feature point pixels in the motion region of the target image and the feature point pixels in the motion region of the reference image are determined respectively. For each feature point pixel in the motion region of the target image or the motion region of the reference image, a descriptor for the feature point pixel is determined based on the pixel gradient of each pixel within the pixel window centered on the feature point pixel. The descriptor is used to characterize the pixel gradient distribution features of each pixel within the pixel window. For each feature point pixel in the motion region of the target image, determine the target feature point pixel in the motion region of the reference image that matches the feature point pixel; determine the temporal denoising coefficient corresponding to the feature point pixel based on the feature distance between the descriptor of the feature point pixel and the descriptor of the target feature point pixel. Based on the temporal denoising coefficients corresponding to the pixels of each feature point in the moving region of the target image, temporal denoising is performed on the moving region of the target image to obtain the temporal denoising result.

[0007] In some embodiments of this application, feature point pixels in the motion region of the target image and feature point pixels in the motion region of the reference image are determined based on the pixel gradients of each pixel in the motion region of the target image and the pixel gradients of each pixel in the motion region of the reference image, respectively, including: For each pixel in the motion region of the target image or reference image, obtain the pixel gradient of each pixel within a pixel window centered on the pixel. Based on the pixel gradient of each pixel, construct the pixel gradient matrix and calculate the response value of the pixel gradient matrix; If the response value is greater than the preset threshold, the pixel will be used as a feature point pixel.

[0008] In some embodiments of this application, the feature point pixels in the motion region of the target image or the feature point pixels in the motion region of the reference image include, but are not limited to, any one or more of the following: edge point pixels, corner point pixels, and texture point pixels.

[0009] In some embodiments of this application, for each feature point pixel in the motion region of the target image, determining a target feature point pixel in the motion region of the reference image that matches the feature point pixel includes: For each feature point pixel in the motion region of the target image, determine the descriptor of the feature point pixel and the feature distance between the feature point pixel and each feature point pixel in the motion region of the reference image; The feature point pixel with the smallest feature distance to the descriptor of the feature point pixel in the motion region of the reference image is taken as the target feature point pixel to be matched with the feature point pixel.

[0010] In some embodiments of this application, the temporal denoising coefficient corresponding to the feature point pixel is determined based on the feature distance between the descriptor of the feature point pixel and the descriptor of the target feature point pixel, including: Based on the preset lower threshold and the preset upper threshold of feature distance, the feature distance between the descriptor of the feature point pixel and the descriptor of the target feature point pixel is mapped to obtain the temporal denoising coefficient corresponding to the feature point pixel.

[0011] In some embodiments of this application, temporal denoising is performed on the moving region of the target image based on the temporal denoising coefficients corresponding to each feature point pixel in the moving region of the target image to obtain a temporal denoising result, including: Based on the temporal noise reduction coefficients corresponding to each feature point pixel in the motion region of the target image, spatial interpolation is performed to obtain the temporal noise reduction coefficients corresponding to each pixel in the motion region of the target image. For each pixel in the motion region of the target image, the temporal denoising coefficient corresponding to the pixel is used as the first weight corresponding to the pixel value; The difference between 1 and the first weight is used as the second weight corresponding to the pixel value of the target pixel with the same pixel position in the reference image; Based on the first weight and the second weight, a weighted summation operation is performed on the pixel value of the pixel and the pixel value of the target pixel to obtain the temporal domain noise reduction value of the pixel; The temporal denoising result is generated based on the temporal denoising value of each pixel in the motion region of the target image.

[0012] In some embodiments of this application, the temporal denoising result is composed of the temporal denoising value of each pixel in the motion region of the target image, and the spatial denoising result corresponding to the motion region is composed of the spatial denoising value of each pixel in the motion region of the target image. The temporal and spatial domain denoising results are fused to obtain the denoised output of the moving region in the target image, including: For each pixel in the motion region of the target image, the fusion coefficient of the pixel is determined based on any one or more of the temporal denoising coefficient corresponding to the pixel, the signal-to-noise characteristic parameters of the pixel, and the inter-frame motion intensity of the pixel. The fusion coefficient is used as the third weight corresponding to the spatial noise reduction value of the pixel; The difference between 1 and the third weight is used as the fourth weight corresponding to the temporal denoising value of the pixel; Based on the third and fourth weights, a weighted summation operation is performed on the spatial domain noise reduction value and the temporal domain noise reduction value of the pixel to obtain the noise reduction output value of the pixel. Based on the noise reduction output value of each pixel in the motion region of the target image, generate the noise reduction output result of the motion region of the target image.

[0013] In some embodiments of this application, the signal-to-noise characteristic parameters of a pixel include any one or more of the local noise intensity, local texture intensity, local edge intensity, and brightness of the pixel.

[0014] According to a second aspect of the embodiments of this application, an image noise reduction apparatus for moving regions is provided, comprising: The acquisition module is used to acquire the target image and the reference image, where the reference image is the previous frame of the target image. The spatial domain noise reduction module is used to perform spatial domain noise reduction on the target image and obtain the spatial domain noise reduction result; The determination module is used to determine the motion region in the target image and the reference image based on the pixel difference between pixels at the same position in the target image and the reference image; The temporal denoising module is used to perform temporal denoising on the motion region of the target image based on the pixel gradient of the pixel in the motion region of the target image and the pixel gradient of the pixel in the motion region of the reference image, and obtain the temporal denoising result. The pixel gradient represents the change of the pixel value in the spatial dimension. The fusion module is used to fuse the spatial denoising results corresponding to the motion regions in the temporal and spatial denoising results to obtain the denoised output results of the motion regions of the target image.

[0015] According to a third aspect of the embodiments of this application, a computer device is provided, the computer device including a processor and a memory, the memory being used to store at least one program, the at least one program being loaded by the processor and executed by any of the above-described image denoising methods for moving regions.

[0016] According to a fourth aspect of the embodiments of this application, a computer-readable storage medium is provided, wherein at least one program is stored in the computer-readable storage medium, and the at least one program is loaded and executed by a processor to implement any of the above-described image denoising methods for moving regions.

[0017] Therefore, in the above method, for the moving region, the terminal device can perform temporal denoising on the moving region of the target image based on the pixel gradient of the pixels in the moving region of the target image and the pixel gradient of the pixels in the moving region of the reference image to obtain the temporal denoising result. Then, the temporal denoising result is fused with the spatial denoising result corresponding to the moving region in the spatial denoising result obtained by spatial denoising of the target image.

[0018] As can be seen, the above method can perform temporal denoising based on pixel gradient information, and can perceive the edge and texture structure represented by pixel gradients during the temporal denoising process, thus improving the temporal denoising effect. Furthermore, the terminal device can fuse the corresponding spatial denoising results based on the temporal denoising results, effectively preserving details of moving areas while denoising, improving image clarity, and further enhancing the image denoising effect. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 A flowchart illustrating an image noise reduction method for moving regions provided in an embodiment of this application; Figure 2 This is a schematic diagram of an image noise reduction device for moving regions provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of a terminal provided in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of a server provided in an embodiment of this application. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0022] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application.

[0023] In this application, the terms "first," "second," etc., are used to distinguish identical or similar items that have essentially the same function. It should be understood that there is no logical or temporal dependency between "first," "second," and "nth," nor does it limit the quantity or execution order. It should also be understood that although the following description uses the terms "first," "second," etc., to describe various objects, these objects should not be limited by the terms.

[0024] These terms are simply used to distinguish one object from another. For example, without departing from the various examples, a first action can be called a second action, and similarly, a second action can be called a first action. Both the first and second actions can be actions, and in some cases, they can be separate and distinct actions.

[0025] "At least one" refers to one or more actions. For example, at least one action can be one action, two actions, three actions, or any integer number of actions greater than or equal to one. "Multiple" refers to two or more actions. For example, multiple actions can be two actions, three actions, or any integer number of actions greater than or equal to two.

[0026] In the field of image processing technology, noise reduction is performed on images to improve their visual quality. When dealing with moving regions in an image, related techniques often employ spatial domain noise reduction. This method only utilizes a single frame of pixels for smoothing, which usually results in image blurring and poor noise reduction performance.

[0027] To address the aforementioned technical problems, this application provides an image denoising method for moving regions, which can improve the image denoising effect.

[0028] Figure 1 This is a flowchart illustrating an image denoising method for moving regions provided in an embodiment of this application. The following is a summary of the process. Figure 1 The method is described in detail below. The executing entity of the method can be a terminal device with image acquisition function, such as a handheld camera, security camera, vehicle camera, smartphone, or tablet computer, or a server that communicates with these terminal devices. The following description only uses the terminal device as the executing entity. The method can include: S101 to S105, as follows.

[0029] S101, acquire the target image and the reference image, where the reference image is the previous frame of the target image.

[0030] The terminal device integrates an image sensor. When the terminal device is running, the image sensor begins to acquire image data (e.g., video stream data), thereby generating a sequential sequence of image frames arranged chronologically. The terminal device can acquire image data in real time. When initiating the image noise reduction process, it selects the current frame to be processed from the image frame sequence as the target image, and simultaneously selects the previous frame as the reference image. For example, when the target image is the second frame in the image frame sequence, the reference image can be the first frame.

[0031] S102, perform spatial domain denoising on the target image to obtain the spatial domain denoising result.

[0032] S103, determine the motion region in the target image and the reference image based on the pixel difference between pixels at the same position in the target image and the reference image.

[0033] The terminal device can perform spatial domain denoising on the target image to obtain a spatial domain denoising result. This result is a two-dimensional pixel matrix with the same size as the target image, composed of the spatial domain denoising value of each pixel in the target image; that is, the image after spatial domain denoising. For example, Gaussian filtering, bilateral filtering, and other methods can be used to achieve spatial domain denoising of the target image. The spatial domain denoising result is shown in the following formula.

[0034]

[0035] in, Represents the coordinates in the target image The pixel value of the previous pixel. This indicates spatial noise reduction operation. Represents the coordinates in the target image Spatial noise reduction value of the upper pixel.

[0036] The terminal device can determine the moving region in the target image and the reference image based on the pixel difference between pixels at the same position in the target image and the reference image, where the moving region is located at the same position in the target image and the reference image.

[0037] Specifically, for each pixel in the target image, the terminal device can first identify the pixel at the same location in the reference image and calculate the pixel difference between the two pixels. This pixel difference is then compared to a preset noise threshold. If the pixel difference is greater than the noise threshold, meaning the pixel value change between the two pixels exceeds the natural fluctuation range of noise, the pixel is determined to be in a moving region. Conversely, if the pixel difference is not greater than the noise threshold, meaning the pixel value change between the two pixels does not exceed the natural fluctuation range of noise, and the pixel value change is solely caused by noise, the pixel is determined to be in a stationary region. The noise threshold can be a pre-set empirical value.

[0038] S104. Based on the pixel gradients of pixels in the motion region of the target image and the pixel gradients of pixels in the motion region of the reference image, temporal domain denoising is performed on the motion region of the target image to obtain the temporal domain denoising result. The pixel gradient represents the change of pixel value in the spatial dimension.

[0039] S105, fuse the temporal domain denoising result and the spatial domain denoising result corresponding to the motion region in the spatial domain denoising result to obtain the denoising output result of the motion region of the target image.

[0040] For moving regions, using only spatial domain denoising can easily lead to image blurring and loss of detail. Therefore, in this embodiment, pixel gradient information can be introduced when denoising moving regions. Adaptive temporal domain denoising is first performed in the moving region, and then the temporal and spatial domain denoising results are adaptively fused. This fully leverages the complementary advantages of thorough temporal denoising and structural fidelity preservation in spatial domain denoising, overcoming the limitations of a single denoising method and achieving a precise balance between extreme denoising and detail preservation in moving regions.

[0041] Specifically, the terminal device can perform temporal denoising on the motion region of the target image based on the pixel gradient of the pixels in the motion region of the target image and the pixel gradient of the pixels in the motion region of the reference image, and obtain the temporal denoising result.

[0042] Pixel gradients characterize the variation of pixel values ​​across spatial dimensions (horizontal, vertical, etc.), reflecting the degree of abrupt changes in the local structure of an image. Smaller pixel gradients indicate smoother pixel value changes in the neighborhood, placing the pixel in a flat region (low-frequency region), allowing for strong filtering using temporal information. Larger pixel gradients indicate more drastic pixel value changes in the neighborhood, placing the pixel in an edge region (high-frequency texture region), minimizing the use of temporal information and preventing motion ghosting, thus achieving adaptive temporal noise reduction based on pixel gradients.

[0043] Then, the terminal device can fuse the temporal denoising results and the spatial denoising results corresponding to the motion regions in the spatial denoising results to obtain the denoising output results of the motion regions of the target image.

[0044] Specifically, the terminal device can adaptively fuse the two noise reduction results. If the pixel is in a flat area and is severely affected by noise, the proportion of the temporal noise reduction result can be increased and the proportion of the spatial noise reduction result can be decreased during the adaptive fusion process, thereby making full use of the advantages of temporal noise reduction to achieve powerful noise reduction. If the pixel is in an edge area and the inter-frame motion intensity is large, the proportion of the temporal noise reduction result can be decreased and the proportion of the spatial noise reduction result can be increased during the adaptive fusion process to suppress motion ghosting.

[0045] Therefore, in the above method, for the moving region, the terminal device can perform temporal denoising on the moving region of the target image based on the pixel gradient of the pixels in the moving region of the target image and the pixel gradient of the pixels in the moving region of the reference image to obtain the temporal denoising result. Then, the temporal denoising result is fused with the spatial denoising result corresponding to the moving region in the spatial denoising result obtained by spatial denoising of the target image.

[0046] As can be seen, the above method can perform temporal denoising based on pixel gradient information, and can perceive the edge and texture structure represented by pixel gradients during the temporal denoising process, thus improving the temporal denoising effect. Furthermore, the terminal device can fuse the corresponding spatial denoising results based on the temporal denoising results, effectively preserving details of moving areas while denoising, improving image clarity, and further enhancing the image denoising effect.

[0047] In some embodiments of this application, S104 is involved. The terminal device can perform temporal denoising on the motion region of the target image based on the pixel gradient of the pixel in the motion region of the target image and the pixel gradient of the pixel in the motion region of the reference image to obtain a temporal denoising result. S104 may include S1041 to S1044, as follows.

[0048] S1041, based on the pixel gradient of each pixel in the motion region of the target image and the pixel gradient of each pixel in the motion region of the reference image, determine the feature point pixels in the motion region of the target image and the feature point pixels in the motion region of the reference image, respectively.

[0049] The terminal device can determine the feature point pixels in the moving region based on the pixel gradient of each pixel in the moving region of the target image. Furthermore, the terminal device can determine the feature point pixels in the moving region based on the pixel gradient of each pixel in the moving region of the reference image.

[0050] A feature point pixel is a pixel in the moving region of the target image or the moving region of the reference image, or a set of pixels in a very small neighborhood. It is a landmark location where the spatial structure of a local region changes abruptly, representing the most recognizable local feature in the image. Feature points pixels can usually be located at the turning point of the outline of a moving object, the extreme point of the boundary, or the intersection of local textures.

[0051] S1042, for each feature point pixel in the motion region of the target image or the motion region of the reference image, a descriptor for the feature point pixel is determined based on the pixel gradient of each pixel within the pixel window centered on the feature point pixel. The descriptor is used to characterize the pixel gradient distribution features of each pixel within the pixel window.

[0052] For each feature point pixel in the motion region of the target image or the motion region of the reference image, the terminal device can first construct a pixel window centered on that feature point pixel. This pixel window is used to obtain the pixel gradient around the pixel. The size of the pixel window can be 7×7, 8×8, etc.

[0053] Then, based on the pixel gradients of each pixel within the pixel window, a descriptor for the feature point pixel is determined. The descriptor is a quantized mathematical vector or feature array that characterizes the pixel gradient distribution features of each pixel within the pixel window. For example, if a feature point pixel is a vertical edge, its descriptor will reflect that the horizontal gradient of the feature point pixel is large, while the vertical gradient is small. As another example, if a feature point pixel is a corner point, its descriptor will reflect that the feature point pixel has significant gradients in multiple directions.

[0054] The descriptors in the embodiments of this application can be binary descriptors or floating-point descriptors, and are not limited to descriptors generated by methods such as ORB, BRIEF, BRISK, SIFT, SURF, neural network models, etc., which will not be described in detail here.

[0055] S1043, for each feature point pixel in the moving region of the target image, determine the target feature point pixel in the moving region of the reference image that matches the feature point pixel; determine the temporal denoising coefficient corresponding to the feature point pixel based on the feature distance between the descriptor of the feature point pixel and the descriptor of the target feature point pixel.

[0056] After determining each feature point pixel, for each feature point pixel in the moving region of the target image, the terminal device can first determine the target feature point pixel in the moving region of the reference image that matches the feature point pixel. The target feature point pixel can be the corresponding point in the reference image that is spatially adjacent to the feature point pixel and has the most similar local spatial structure.

[0057] Then, the terminal device can determine the temporal denoising coefficient corresponding to the feature point pixel based on the feature distance between the descriptor of the feature point pixel and the descriptor of the target feature point pixel. The temporal denoising coefficient can determine the proportion of the target image in the temporal denoising process.

[0058] The smaller the feature distance, the more similar the two feature point pixels are in local spatial structure. The smaller the temporal denoising coefficient corresponding to the feature point pixel, the more information from the reference image can be used to achieve image denoising when performing temporal denoising.

[0059] Conversely, the larger the feature distance, the greater the difference between the two feature point pixels in the local spatial structure, and the larger the temporal denoising coefficient corresponding to the feature point pixel. In temporal denoising, more information of the target image can be used to avoid ghosting, trailing and other phenomena.

[0060] In the above method, feature point pixels can be determined based on pixel gradients, and temporal denoising coefficients can be determined based on the feature distance between the feature point pixels and the target feature point pixels for temporal denoising. By using pixel gradients for temporal denoising, the absolute brightness value of the pixels can be eliminated, reducing the impact of overall image brightness changes, exposure changes, or black level shifts on feature distance calculation. This ensures the accuracy of temporal matching evaluation under complex lighting fluctuations, avoids misjudgment of temporal denoising coefficients due to ambient light changes, and improves the effect of temporal denoising.

[0061] S1044, Based on the temporal denoising coefficients corresponding to the feature point pixels in the moving region of the target image, temporal denoising is performed on the moving region of the target image to obtain the temporal denoising result.

[0062] After determining the temporal denoising coefficients corresponding to each feature point pixel in the motion region of the target image, the terminal device can perform temporal denoising on the motion region of the target image based on each temporal denoising coefficient. By performing temporal fusion on the motion region of the reference image and the motion region of the target image, the temporal denoising result is obtained.

[0063] In some embodiments of this application, S1041 is involved. The terminal device can determine the feature point pixels in the motion region of the target image and the feature point pixels in the motion region of the reference image based on the pixel gradient of each pixel in the motion region of the target image and the pixel gradient of each pixel in the motion region of the reference image, respectively. S1041 may include S10411 to S10413, as follows: S10411, for each pixel in the motion region of the target image or reference image, obtain the pixel gradient of each pixel within a pixel window centered on the pixel.

[0064] S10412: Based on the pixel gradient of each pixel, construct the pixel gradient matrix and calculate the response value of the pixel gradient matrix.

[0065] S10413 If the response value is greater than the preset threshold, the pixel is used as a feature point pixel.

[0066] The aforementioned pixel window is a detection window used to perceive extremely small local spatial structures within pixels. The size of this pixel window can be 2×2, 3×3, etc. For each pixel in the moving region of the target image or reference image, the terminal device can acquire the pixel gradient of each pixel within the pixel window centered on that pixel.

[0067] Furthermore, based on the pixel gradients of each pixel, a pixel gradient matrix can be constructed, and its response value can be calculated. The pixel gradient matrix is ​​a structure tensor matrix composed of the products of the gradients of each pixel within the pixel window in the horizontal and vertical directions. It reflects the gradient distribution of the image within the pixel window in two orthogonal directions. The response value of the pixel gradient matrix is ​​a scalar that measures the drastic change in pixel value when the pixel window shifts in any direction.

[0068] The terminal device can compare the response value with a preset threshold (i.e., a pre-set empirical value). The preset threshold is used to filter out small gradient fluctuations and unidirectional ordinary edges caused by image noise, and to select real corner points or texture intersections with significant multi-directional structural changes. If the response value is greater than the preset threshold, the pixel is determined to have sufficient spatial structural significance and is marked as a feature point pixel.

[0069] In other embodiments of this application, methods such as FAST and ORB can also be used to determine feature point pixels, which will not be described in detail here.

[0070] In some embodiments of this application, the feature point pixels in the motion region of the target image or the feature point pixels in the motion region of the reference image can be pixels with pixel gradient abrupt changes, such as edge pixels, corner pixels, and texture pixels.

[0071] In some embodiments of this application, S1043 is involved, in which, for each feature point pixel in the motion region of the target image, the terminal device can determine the target feature point pixel in the motion region of the reference image that matches the feature point pixel. S1043 may include S10431 and S10432, as follows.

[0072] S10431, for each feature point pixel in the motion region of the target image, determine the descriptor of the feature point pixel and the feature distance between the feature point pixel and each feature point pixel in the motion region of the reference image.

[0073] S10432, the feature point pixel with the smallest feature distance to the descriptor of the feature point pixel in the motion region of the reference image is taken as the target feature point pixel to be matched with the feature point pixel.

[0074] For each feature point pixel in the motion region of the target image, the terminal device can determine the feature distance between the descriptor of the feature point pixel and each feature point pixel in the motion region of the reference image. Then, the feature point pixel in the motion region of the reference image with the smallest feature distance to the descriptor of the feature point pixel can be used as the target feature point pixel that matches the feature point pixel.

[0075] In this embodiment, the above-mentioned characteristic distances can be determined by methods such as Hamming distance, Euclidean distance, Manhattan distance, cosine distance, and cross-correlation response, which will not be described in detail here.

[0076] In some embodiments of this application, S1043 is involved, in which the terminal device determines the temporal denoising coefficient corresponding to the feature point pixel based on the feature distance between the descriptor of the feature point pixel and the descriptor of the target feature point pixel. S1043 may specifically include: Two feature distance thresholds can be preset: a lower feature distance threshold and an upper feature distance threshold. If the feature distance between the descriptor of a feature pixel and the descriptor of its corresponding target feature pixel is less than the lower feature distance threshold, the local spatial structures of the two feature pixels are considered to be consistent. This allows the temporal denoising coefficient corresponding to the feature pixel to be minimized, thereby maximizing the use of information from the reference image for temporal denoising.

[0077] If the feature distance between the descriptor of a feature point pixel and the descriptor of its corresponding target feature point pixel is greater than the upper limit threshold of the feature distance, it is determined that the local spatial structures of the two feature point pixels are completely different. The temporal denoising coefficient corresponding to the feature point pixel can be increased to the maximum to make the most of the information of the target image for temporal denoising.

[0078] Specifically, the terminal device can perform feature distance mapping between the descriptor of the feature point pixel and the descriptor of the target feature point pixel based on a preset lower threshold and a preset upper threshold for feature distance, to obtain the temporal denoising coefficient corresponding to the feature point pixel, as shown in the following formula.

[0079]

[0080] in, This represents the lower limit threshold of the feature distance. This represents the upper limit threshold of the feature distance. This represents the temporal denoising coefficient corresponding to feature point pixel k. This represents the feature distance between feature point pixel k and its corresponding target feature point pixel. This represents a limiting mapping function used to restrict the values ​​of the time-domain denoising coefficients to between 0.0 and 1.0. Less than , The value is 0.0. Greater than , The value is 1.0, if In and between, along with The value increases linearly between 0.0 and 1.0.

[0081] In some embodiments of this application, S1044 is involved. The terminal device can perform temporal denoising on the moving region of the target image according to the temporal denoising coefficients corresponding to each feature point pixel in the moving region of the target image to obtain the temporal denoising result. S1044 may include: S10441 to S10445, as follows.

[0082] S10441, based on the temporal denoising coefficients corresponding to each feature point pixel in the motion region of the target image, perform spatial interpolation processing to obtain the temporal denoising coefficients corresponding to each pixel in the motion region of the target image.

[0083] Terminal devices can use spatial interpolation to diffuse the temporal denoising coefficients corresponding to each feature point pixel into the temporal denoising coefficients corresponding to each pixel in the moving region of the target image. For example, nearest neighbor diffusion, bilinear interpolation, guided interpolation based on image edges, block-level diffusion, morphological dilation, and edge-preserving diffusion methods combined with image edges can be used, etc., which are not specifically limited here.

[0084] S10442, for each pixel in the moving region of the target image, the temporal denoising coefficient corresponding to the pixel is used as the first weight corresponding to the pixel value.

[0085] S10443, the difference between 1 and the first weight is used as the second weight corresponding to the pixel value of the target pixel with the same pixel position in the reference image.

[0086] S10444: Based on the first weight and the second weight, perform a weighted summation operation on the pixel value of the pixel and the pixel value of the target pixel to obtain the temporal domain noise reduction value of the pixel.

[0087] After determining the temporal denoising coefficients for each pixel in the moving region of the target image, temporal denoising can be performed. Specifically, for each pixel in the moving region of the target image, the temporal denoising coefficient corresponding to that pixel can be used as the first weight corresponding to the pixel value of that pixel, and the difference between 1 and the first weight can be used as the second weight corresponding to the pixel value of the target pixel in the reference image at the same position as that pixel. Then, based on the first weight and the second weight, a weighted summation operation can be performed on the pixel value of the target pixel and the pixel value of the target pixel to obtain the temporal denoising value of that pixel, as shown in the following formula.

[0088]

[0089] in, Represents the coordinates in the target image The pixel value of the previous pixel. Represents the coordinates in the target image Temporal noise reduction value of the upper pixel, Represents the coordinates in the target image The temporal denoising coefficient (i.e., the first weight) corresponding to the upper pixel. Represents coordinates in the reference image The pixel value of the previous pixel. Represents coordinates in the reference image The second weight corresponding to the pixel value of the previous pixel.

[0090] S10445, Generate temporal denoising results based on the temporal denoising values ​​of each pixel in the motion region of the target image.

[0091] Then, based on the temporal denoising value corresponding to each pixel in the moving region of the target image, the temporal denoising values ​​corresponding to each pixel can be combined and arranged sequentially according to the arrangement order of each pixel in the moving region of the target image to generate a temporal denoising result. This temporal denoising result is a two-dimensional pixel matrix with the same size as the moving region of the target image, that is, the image after temporal denoising.

[0092] In some embodiments of this application, the above-mentioned temporal denoising result is composed of the temporal denoising value of each pixel in the motion region of the target image, and the spatial denoising result corresponding to the motion region is composed of the spatial denoising value of each pixel in the motion region of the target image. The spatial denoising value is the pixel value obtained after spatial denoising processing.

[0093] In connection with S105, the terminal device can fuse the spatial denoising results corresponding to the motion region in the temporal denoising results and the spatial denoising results to obtain the denoising output result of the motion region of the target image. S105 can include: S1051 to S1055, as follows.

[0094] S1051, for each pixel in the moving region of the target image, determine the pixel fusion coefficient based on any one or more of the following: the temporal denoising coefficient corresponding to the pixel, the signal-to-noise characteristic parameter of the pixel, and the inter-frame motion intensity of the pixel.

[0095] The signal-to-noise ratio (SNR) parameters of a pixel can be used to evaluate the pixel's own signal properties (brightness, texture, edges, etc.) and the level of environmental noise. The inter-frame motion intensity of a pixel can characterize the magnitude of the change in pixel value between the pixel and a pixel at the same position in the reference image, and is used to evaluate the stability of the pixel.

[0096] For each pixel in the moving region of the target image, the terminal device can determine the fusion coefficient of the pixel based on any one or more of the temporal denoising coefficient, the signal-to-noise characteristic parameters of the pixel, and the inter-frame motion intensity of the pixel. The fusion coefficient can determine the proportion of the spatial denoising result in the fusion process.

[0097] The fusion coefficient is positively correlated with the temporal denoising coefficient. As the temporal denoising coefficient increases, the reliability of the reference image decreases. Therefore, the fusion coefficient can be increased to increase the proportion of spatial denoising results and decrease the proportion of temporal denoising results during the fusion process, thereby suppressing ghosting in moving regions. Alternatively, the temporal denoising coefficient can be directly used as the fusion coefficient.

[0098] Inter-frame motion intensity is positively correlated with the fusion coefficient. As the inter-frame motion intensity increases, indicating that the reference image is unreliable, the fusion coefficient can be increased to increase the proportion of spatial denoising results and decrease the proportion of temporal denoising results during the fusion process, thereby suppressing ghosting in moving regions. Alternatively, the temporal denoising coefficient can be directly used as the fusion coefficient.

[0099] If the signal-to-noise characteristic parameter indicates that the current noise is large, that is, the current pixel is severely affected by random noise, the fusion coefficient can be reduced to decrease the proportion of spatial domain noise reduction results and increase the proportion of temporal domain noise reduction results during the fusion process. This avoids the patch / painting effect that is easily generated by strong spatial domain filtering and uses the average depth of multiple frames in the temporal domain to suppress strong noise.

[0100] If the signal-to-noise characteristic parameters indicate strong edges and textures, that is, the current region has rich details or abrupt structural changes, the fusion coefficient can be increased to increase the proportion of spatial domain noise reduction results and decrease the proportion of temporal domain noise reduction results during the fusion process. This avoids edge trailing and texture flickering caused by temporal registration errors and maintains sharp contours by utilizing the spatial edge preservation characteristics.

[0101] If the signal-to-noise ratio characteristic parameter indicates low brightness, that is, the current pixel is in a dark area and the sensor signal-to-noise ratio is extremely low, then the fusion coefficient can be reduced to reduce the proportion of spatial domain noise reduction results and increase the proportion of temporal domain noise reduction results during the fusion process, thereby compensating for the physical disadvantage of dark areas and eliminating coarse noise in dark fields by relying on temporal integration.

[0102] S1052 uses the fusion coefficient as the third weight corresponding to the spatial noise reduction value of the pixel.

[0103] S1053, the difference between 1 and the third weight is used as the fourth weight corresponding to the temporal domain noise reduction value of the pixel.

[0104] S1054 performs a weighted summation operation on the spatial domain noise reduction value and the temporal domain noise reduction value of the pixel according to the third weight and the fourth weight, to obtain the noise reduction output value of the pixel.

[0105] The noise reduction output value of a pixel is specifically calculated using the following formula.

[0106]

[0107] in, Represents the coordinates in the target image Spatial noise reduction value of the upper pixel, Represents the coordinates in the target image Temporal noise reduction value of the upper pixel, Represents the coordinates in the target image The fusion coefficient of the upper pixel (i.e., the third weight). Indicates the fourth weight. Indicates the coordinates in the target image as The noise reduction output value corresponding to each pixel.

[0108] S1055, based on the noise reduction output value of each pixel in the motion region of the target image, generates the noise reduction output result of the motion region of the target image.

[0109] After determining the noise reduction output value of each pixel in the motion region of the target image, the noise reduction output values ​​corresponding to each pixel can be combined and arranged sequentially according to the arrangement order of each pixel in the motion region of the target image to generate the noise reduction output result of the motion region of the target image. The noise reduction output result is a two-dimensional pixel matrix with the same size as the motion region of the target image, that is, the final noise-reduced image.

[0110] As can be seen, the above method can adaptively adjust the ratio of temporal and spatial denoising results in the motion region during the fusion process. This allows for an increase in the proportion of spatial denoising results in areas where the reference image is unreliable, edges are present, or textures are strong, in order to suppress ghosting and preserve detail sharpness. In areas with severe noise interference or low brightness, more temporal denoising effects are retained for deep denoising and to avoid spatial patch effects, thereby reducing ghosting and motion blur. At the same time, it improves the problems of uneven noise morphology, blurred texture, or noise residue in the motion region.

[0111] In some embodiments of this application, the signal-to-noise characteristic parameters of a pixel may include any one or more of the local noise intensity, local texture intensity, local edge intensity, and brightness of the pixel.

[0112] Among them, the local noise intensity corresponding to a pixel can reflect the degree of random noise interference to the current pixel, the local texture intensity corresponding to a pixel can reflect the richness of detail and the severity of edge changes in the area where the current pixel is located, the local edge intensity corresponding to a pixel can reflect the real physical boundary of the area where the current pixel is located, and the brightness of a pixel can reflect the signal strength of the current pixel and the degree of influence of dark current noise.

[0113] In some embodiments of this application, the above-mentioned fusion parameters can be calculated by combining the above-mentioned signal-to-noise characteristic parameters, temporal denoising coefficients, and inter-frame motion intensity, as shown in the following formula.

[0114]

[0115] in, Represents the coordinates in the target image The fusion coefficient of the upper pixel, This represents the amplitude limiting mapping function, used to restrict the values ​​of the fusion coefficients to between 0.0 and 1.0. Represents the coordinates in the target image The temporal noise reduction coefficient corresponding to the upper pixel, express The corresponding weights Represents the coordinates in the target image The local noise intensity corresponding to the upper pixel, express The corresponding weights Represents the coordinates in the target image The local texture intensity corresponding to the previous pixel. express The corresponding weights Represents the coordinates in the target image Brightness of the upper pixel, express The corresponding weights Represents the coordinates in the target image The local edge intensity corresponding to the upper pixel, express The corresponding weights Represents the coordinates in the target image Inter-frame motion intensity of the upper pixel, express The corresponding weights, where the weights for each item can be empirical constants pre-calibrated based on the test.

[0116] In some embodiments of this application, in order to improve the image denoising effect, a multi-frame iterative denoising method can be used to perform image denoising, thereby making full use of the correlation between multiple adjacent frames, avoiding insufficient denoising, and making the denoised image clearer, smoother, and more visually stable. Specifically, after the denoising of the current processing frame (i.e., the target image) is completed and the denoising output result is obtained, the denoising output result can be used as the reference image for the next processing frame, so as to facilitate subsequent image denoising.

[0117] In some embodiments of this application, when performing image denoising on a target image, the static regions in the target image other than the moving regions can be denoised only in the temporal domain to obtain the denoised output result corresponding to the static regions.

[0118] As can be seen, the above method can perform temporal denoising based on pixel gradient information, and can perceive the edge and texture structure represented by pixel gradients during the temporal denoising process, thus improving the temporal denoising effect. Furthermore, the terminal device can fuse the corresponding spatial denoising results based on the temporal denoising results, effectively preserving details of moving areas while denoising, improving image clarity, and further enhancing the image denoising effect.

[0119] In addition, the above method can adaptively adjust the ratio of temporal and spatial denoising results in the motion region during the fusion process. This allows for an increase in the proportion of spatial denoising results in areas where the reference image is unreliable, edges are present, or textures are strong, in order to suppress ghosting and preserve detail sharpness. In areas with severe noise interference or low brightness, more temporal denoising effects are retained for deep denoising and to avoid spatial patch effects, thereby reducing ghosting and motion blur. At the same time, it improves the problems of uneven noise morphology, blurred textures, or noise residue in the motion region.

[0120] Embodiments of this application also provide an image noise reduction device for moving regions, such as... Figure 2 As shown, Figure 2 This is a schematic diagram of an image noise reduction device for moving areas provided in an embodiment of this application. The image noise reduction device 200 for moving areas can be deployed on terminal devices with image acquisition capabilities, such as handheld cameras, security cameras, vehicle cameras, smartphones, and tablets, or it can be deployed on a server communicating with these terminal devices. The image noise reduction device 200 for moving areas can integrate semiconductor logic circuits and control circuits to achieve noise reduction processing of the image. The image noise reduction device 200 for moving areas includes: The acquisition module 201 is used to acquire the target image and the reference image, wherein the reference image is the previous frame image of the target image; The spatial domain noise reduction module 202 is used to perform spatial domain noise reduction on the target image to obtain the spatial domain noise reduction result. The determination module 203 is used to determine the motion region in the target image and the reference image based on the pixel difference between pixels at the same position in the target image and the reference image; The temporal denoising module 204 is used to perform temporal denoising on the motion region of the target image based on the pixel gradient of the pixel in the motion region of the target image and the pixel gradient of the pixel in the motion region of the reference image, so as to obtain the temporal denoising result. The pixel gradient represents the change of the pixel value of the pixel in the spatial dimension. The fusion module 205 is used to fuse the spatial denoising results corresponding to the motion regions in the temporal denoising results and the spatial denoising results to obtain the denoising output results of the motion regions of the target image.

[0121] In some embodiments of this application, the temporal denoising module 204 is specifically used to: determine feature point pixels in the motion region of the target image and feature point pixels in the motion region of the reference image, respectively, based on the pixel gradients of each pixel in the motion region of the target image and the pixel gradients of each pixel in the motion region of the reference image; for each feature point pixel in the motion region of the target image or the motion region of the reference image, determine a descriptor for the feature point pixel based on the pixel gradients of each pixel within a pixel window centered on the feature point pixel, wherein the descriptor is used to characterize the pixel gradient distribution features of each pixel within the pixel window; for each feature point pixel in the motion region of the target image, determine a target feature point pixel in the motion region of the reference image that matches the feature point pixel; determine the temporal denoising coefficient corresponding to the feature point pixel based on the feature distance between the descriptor of the feature point pixel and the descriptor of the target feature point pixel; and perform temporal denoising on the motion region of the target image based on the temporal denoising coefficients corresponding to each feature point pixel in the motion region of the target image to obtain a temporal denoising result.

[0122] In some embodiments of this application, the temporal denoising module 204 is specifically used to: for each pixel in the motion region of the target image or reference image, obtain the pixel gradient of each pixel within a pixel window centered on the pixel; construct a pixel gradient matrix based on the pixel gradient of each pixel and calculate the response value of the pixel gradient matrix; if the response value is greater than a preset threshold, treat the pixel as a feature point pixel.

[0123] In some embodiments of this application, the feature point pixels in the motion region of the target image or the feature point pixels in the motion region of the reference image include, but are not limited to, any one or more of the following: edge point pixels, corner point pixels, and texture point pixels.

[0124] In some embodiments of this application, the temporal denoising module 204 is specifically used to: for each feature point pixel in the motion region of the target image, determine the feature distance between the descriptor of the feature point pixel and each feature point pixel in the motion region of the reference image; and select the feature point pixel in the motion region of the reference image with the smallest feature distance to the descriptor of the feature point pixel as the target feature point pixel that matches the feature point pixel.

[0125] In some embodiments of this application, the temporal denoising module 204 is specifically used to perform feature distance mapping processing between the descriptor of the feature point pixel and the descriptor of the target feature point pixel based on a preset lower feature distance threshold and a preset upper feature distance threshold, so as to obtain the temporal denoising coefficient corresponding to the feature point pixel.

[0126] In some embodiments of this application, the temporal denoising module 204 is specifically used to: perform spatial interpolation processing based on the temporal denoising coefficients corresponding to each feature point pixel in the motion region of the target image to obtain the temporal denoising coefficients corresponding to each pixel in the motion region of the target image; for each pixel in the motion region of the target image, use the temporal denoising coefficient corresponding to the pixel as the first weight corresponding to the pixel value of the pixel; use the difference between 1 and the first weight as the second weight corresponding to the pixel value of the target pixel in the reference image with the same pixel position; perform a weighted summation operation on the pixel value of the pixel and the pixel value of the target pixel based on the first weight and the second weight to obtain the temporal denoising value of the pixel; and generate a temporal denoising result based on the temporal denoising value of each pixel in the motion region of the target image.

[0127] In some embodiments of this application, the temporal denoising result is composed of the temporal denoising value of each pixel in the motion region of the target image, and the spatial denoising result corresponding to the motion region is composed of the spatial denoising value of each pixel in the motion region of the target image. The fusion module 205 is specifically used to: determine the fusion coefficient of each pixel in the motion region of the target image based on any one or more of the temporal denoising coefficient, the signal-to-noise characteristic parameters of the pixel, and the inter-frame motion intensity of the pixel; use the fusion coefficient as the third weight corresponding to the spatial denoising value of the pixel; use the difference between 1 and the third weight as the fourth weight corresponding to the temporal denoising value of the pixel; perform a weighted summation operation on the spatial denoising value and the temporal denoising value of the pixel based on the third weight and the fourth weight to obtain the denoising output value of the pixel; and generate the denoising output result of the motion region of the target image based on the denoising output value of each pixel in the motion region of the target image.

[0128] In some embodiments of this application, the signal-to-noise characteristic parameters of a pixel include any one or more of the local noise intensity, local texture intensity, local edge intensity, and brightness of the pixel.

[0129] Embodiments of this application also provide a computer device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement any of the above-described methods for image denoising of moving regions.

[0130] Taking computer devices as terminals as an example, Figure 3 A schematic diagram of the structure of a terminal provided in an embodiment of this application is shown below. Figure 3 In some embodiments of this application, terminal 300 may be: a smartphone, tablet computer, MP3 player (Moving Picture Experts Group Audio Layer III), MP4 player (Moving Picture Experts Group Audio Layer IV), laptop computer, or desktop computer. Terminal 300 may also be referred to as user equipment, portable terminal, laptop terminal, desktop terminal, or other names.

[0131] Typically, terminal 300 includes a processor 301 and a memory 302.

[0132] In some embodiments of this application, processor 301 may include one or more processing cores, such as a quad-core processor, a penta-core processor, etc. Processor 301 may be implemented using at least one hardware form selected from DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). Processor 301 may also include a main processor and a coprocessor. The main processor, also known as a CPU (Central Processing Unit), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, processor 301 may integrate a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the screen. In some embodiments, processor 301 may also include an AI (Artificial Intelligence) processor, which is used to handle computational operations related to machine learning.

[0133] The memory 302 may include one or more computer-readable storage media, which may be non-transitory. The memory 302 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In some embodiments, the non-transitory computer-readable storage media in the memory 302 are used to store at least one program code, which is executed by the processor 301 to implement the process of terminal execution in the method embodiments of this application.

[0134] In some embodiments, the terminal 300 may also optionally include a peripheral device interface 303 and at least one peripheral device. The processor 301, memory 302, and peripheral device interface 303 can be connected via a bus or signal line. Each peripheral device can be connected to the peripheral device interface 303 via a bus, signal line, or circuit board. Specifically, the peripheral device includes at least one of a display screen 304, a camera assembly 305, an audio circuit 306, and a power supply 307.

[0135] The peripheral device interface 303 can be used to connect at least one I / O (Input / Output) related peripheral device to the processor 301 and the memory 302. In some embodiments, the processor 301, memory 302, and peripheral device interface 303 are integrated on the same chip or circuit board; in some other embodiments, any one or two of the processor 301, memory 302, and peripheral device interface 303 can be implemented on separate chips or circuit boards, and this application embodiment does not limit this.

[0136] Display screen 304 is used to display a UI (User Interface). This UI may include graphics, text, icons, videos, and any combination thereof. When display screen 304 is a touch display screen, it also has the ability to collect touch signals on or above its surface. These touch signals can be input as control signals to processor 301 for processing. In this case, display screen 304 can also be used to provide virtual buttons and / or a virtual keyboard, also known as soft buttons and / or a soft keyboard. In some embodiments, there may be one display screen 304, disposed on the front panel of terminal 300; in other embodiments, there may be at least two display screens, disposed on different surfaces of terminal 300 or in a folded design; in other embodiments, display screen 304 may be a flexible display screen, disposed on a curved or folded surface of terminal 300. Furthermore, display screen 304 may be configured as a non-rectangular, irregular shape, i.e., a non-rectangular screen. Display screen 304 may be made of materials such as LCD (Liquid Crystal Display) or OLED (Organic Light-Emitting Diode).

[0137] The camera assembly 305 is used to acquire images or videos. In some embodiments, the camera assembly 305 includes a front-facing camera and a rear-facing camera. Typically, the front-facing camera is located on the front panel of the terminal, and the rear-facing camera is located on the back of the terminal. In some embodiments, there are at least two rear-facing cameras, which are any one of a main camera, a depth-sensing camera, a wide-angle camera, and a telephoto camera, to achieve background blurring by fusion of the main camera and the depth-sensing camera, panoramic shooting by fusion of the main camera and the wide-angle camera, VR (Virtual Reality) shooting, or other fusion shooting functions. In some embodiments, the camera assembly 305 may also include a flash. The flash can be a single-color temperature flash or a dual-color temperature flash. A dual-color temperature flash is a combination of a warm-light flash and a cool-light flash, which can be used for light compensation at different color temperatures.

[0138] The audio circuit 306 may include a microphone and a speaker. The microphone is used to collect sound waves from the user and the environment, and convert the sound waves into electrical signals which are then input to the processor 301 for processing. For stereo sound acquisition or noise reduction purposes, multiple microphones may be used, each positioned at a different location on the terminal 300. The microphone may also be an array microphone or an omnidirectional microphone. The speaker is used to convert the electrical signals from the processor 301 into sound waves. The speaker may be a conventional diaphragm speaker or a piezoelectric ceramic speaker. When the speaker is a piezoelectric ceramic speaker, it can convert electrical signals not only into audible sound waves but also into inaudible sound waves for purposes such as distance measurement. In some embodiments, the audio circuit 306 may also include a headphone jack.

[0139] Power supply 307 is used to power the various components in terminal 300. Power supply 307 can be AC ​​power, DC power, a disposable battery, or a rechargeable battery. When power supply 307 includes a rechargeable battery, the rechargeable battery can support wired or wireless charging. The rechargeable battery can also be used to support fast charging technology.

[0140] Those skilled in the art will understand that Figure 3 The structure shown does not constitute a limitation on terminal 300, and may include more or fewer components than shown, or combine certain components, or use different component arrangements.

[0141] Taking computer equipment as a server as an example, Figure 4 This is a schematic diagram of a server structure provided in an embodiment of this application. The server 400 can vary significantly due to different configurations or performance. It may include one or more processors 401 (Central Processing Units, CPUs) and one or more memories 402. The one or more memories 402 store at least one computer program, which is loaded and executed by the one or more processors 401 to implement the aforementioned image noise reduction method. Of course, the server 400 may also have wired or wireless network interfaces, a keyboard, and input / output interfaces for input and output. The server 400 may also include other components for implementing device functions, which will not be elaborated upon here.

[0142] Embodiments of this application also provide a computer-readable storage medium, which includes a stored computer program, wherein the computer program, when running, controls the device where the computer-readable storage medium is located to perform the method described above. Optionally, the computer-readable storage medium may be read-only memory (ROM), random access memory (RAM), compact-disc read-only memory (CD-ROM), magnetic tape, floppy disk, and optical data storage device, etc.

[0143] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.

[0144] The above description is merely an optional embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for image denoising targeting moving regions, characterized in that, include: Acquire a target image and a reference image, wherein the reference image is the previous frame image of the target image; Spatial domain denoising is performed on the target image to obtain the spatial domain denoising result; Based on the pixel difference between pixels at the same position in the target image and the reference image, the motion region in the target image and the reference image is determined; Based on the pixel gradients of pixels in the motion region of the target image and the pixel gradients of pixels in the motion region of the reference image, temporal denoising is performed on the motion region of the target image to obtain a temporal denoising result. The pixel gradient represents the change of pixel value in the spatial dimension. The temporal denoising result and the spatial denoising result corresponding to the motion region in the spatial denoising result are fused to obtain the denoising output result of the motion region of the target image.

2. The method according to claim 1, characterized in that, Based on the pixel gradients of pixels in the motion region of the target image and the pixel gradients of pixels in the motion region of the reference image, temporal denoising is performed on the motion region of the target image to obtain a temporal denoising result, including: Based on the pixel gradients of each pixel in the motion region of the target image and the pixel gradients of each pixel in the motion region of the reference image, feature point pixels in the motion region of the target image and feature point pixels in the motion region of the reference image are determined respectively. For each feature point pixel in the motion region of the target image or the motion region of the reference image, a descriptor for the feature point pixel is determined based on the pixel gradient of each pixel within a pixel window centered on the feature point pixel. The descriptor is used to characterize the pixel gradient distribution features of each pixel within the pixel window. For each feature point pixel in the motion region of the target image, a target feature point pixel in the motion region of the reference image that matches the feature point pixel is determined; based on the feature distance between the descriptor of the feature point pixel and the descriptor of the target feature point pixel, the temporal denoising coefficient corresponding to the feature point pixel is determined; Based on the temporal denoising coefficients corresponding to each feature point pixel in the motion region of the target image, temporal denoising is performed on the motion region of the target image to obtain the temporal denoising result.

3. The method according to claim 2, characterized in that, Based on the pixel gradients of each pixel in the motion region of the target image and the pixel gradients of each pixel in the motion region of the reference image, feature point pixels in the motion region of the target image and feature point pixels in the motion region of the reference image are determined, including: For each pixel in the motion region of the target image or the reference image, obtain the pixel gradient of each pixel within a pixel window centered on the pixel; Based on the pixel gradient of each pixel, construct a pixel gradient matrix and calculate the response value of the pixel gradient matrix; If the response value is greater than a preset threshold, the pixel is used as a feature point pixel.

4. The method according to claim 2, characterized in that, Feature point pixels in the motion region of the target image or feature point pixels in the motion region of the reference image include, but are not limited to, any one or more of the following: edge point pixels, corner point pixels, and texture point pixels.

5. The method according to claim 2, characterized in that, For each feature point pixel in the motion region of the target image, determining the target feature point pixel in the motion region of the reference image that matches the feature point pixel includes: For each feature point pixel in the motion region of the target image, determine the descriptor of the feature point pixel and the feature distance between the feature point pixel and each feature point pixel in the motion region of the reference image; The feature point pixel with the smallest feature distance to the descriptor of the feature point pixel in the motion region of the reference image is taken as the target feature point pixel to match the feature point pixel.

6. The method according to claim 2, characterized in that, Based on the feature distance between the descriptor of the feature point pixel and the descriptor of the target feature point pixel, the temporal denoising coefficient corresponding to the feature point pixel is determined, including: Based on a preset lower threshold and a preset upper threshold for feature distance, the feature distance between the descriptor of the feature point pixel and the descriptor of the target feature point pixel is mapped to obtain the temporal denoising coefficient corresponding to the feature point pixel.

7. The method according to claim 2, characterized in that, Based on the temporal denoising coefficients corresponding to each feature point pixel in the motion region of the target image, temporal denoising is performed on the motion region of the target image to obtain the temporal denoising result, including: Based on the temporal noise reduction coefficients corresponding to each feature point pixel in the motion region of the target image, spatial interpolation is performed to obtain the temporal noise reduction coefficients corresponding to each pixel in the motion region of the target image. For each pixel in the motion region of the target image, the temporal denoising coefficient corresponding to the pixel is used as the first weight corresponding to the pixel value; The difference between 1 and the first weight is used as the second weight corresponding to the pixel value of the target pixel in the reference image that is at the same pixel position as the pixel. Based on the first weight and the second weight, a weighted summation operation is performed on the pixel value of the pixel and the pixel value of the target pixel to obtain the temporal domain noise reduction value of the pixel; A temporal denoising result is generated based on the temporal denoising value of each pixel in the motion region of the target image.

8. The method according to claim 7, characterized in that, The temporal denoising result is composed of the temporal denoising value of each pixel in the motion region of the target image, and the spatial denoising result corresponding to the motion region is composed of the spatial denoising value of each pixel in the motion region of the target image. The temporal denoising result and the spatial denoising result corresponding to the motion region in the spatial denoising result are fused to obtain the denoised output result of the motion region of the target image, including: For each pixel in the motion region of the target image, the fusion coefficient of the pixel is determined based on any one or more of the temporal denoising coefficient corresponding to the pixel, the signal-to-noise feature parameters of the pixel, and the inter-frame motion intensity of the pixel. The fusion coefficient is used as the third weight corresponding to the spatial noise reduction value of the pixel; The difference between 1 and the third weight is used as the fourth weight corresponding to the temporal domain noise reduction value of the pixel; Based on the third weight and the fourth weight, a weighted summation operation is performed on the spatial domain noise reduction value and the temporal domain noise reduction value of the pixel to obtain the noise reduction output value of the pixel. Based on the noise reduction output value of each pixel in the motion region of the target image, a noise reduction output result for the motion region of the target image is generated.

9. The method according to claim 8, characterized in that, The signal-to-noise characteristic parameters of the pixel include any one or more of the local noise intensity, local texture intensity, local edge intensity, and brightness of the pixel.

10. An image noise reduction device for moving regions, characterized in that, include: The acquisition module is used to acquire a target image and a reference image, wherein the reference image is the previous frame image of the target image; The spatial domain noise reduction module is used to perform spatial domain noise reduction on the target image to obtain the spatial domain noise reduction result. The determining module is used to determine the motion region in the target image and the reference image based on the pixel difference between pixels at the same position in the target image and the reference image; The temporal denoising module is used to perform temporal denoising on the motion region of the target image based on the pixel gradient of the pixel in the motion region of the target image and the pixel gradient of the pixel in the motion region of the reference image, so as to obtain the temporal denoising result. The pixel gradient represents the change of the pixel value of the pixel in the spatial dimension. The fusion module is used to fuse the temporal denoising result and the spatial denoising result corresponding to the motion region in the spatial denoising result to obtain the denoising output result of the motion region of the target image.

Images