Image noise reduction method, device, equipment and storage medium

Abstract

The embodiment of the present application provides an image denoising method, apparatus, device and storage medium, wherein the image to be processed is input into an image denoising model for denoising, and an initial denoised image corresponding to the image to be processed output by the image denoising model is obtained; the image to be processed is input into an annotation model for pixel annotation, and pixel information corresponding to the image to be processed output by the annotation model is obtained; based on the pixel information and the initial denoised image, a target denoised image corresponding to the image to be processed is obtained. Image denoising is achieved by combining an image denoising model and a pixel-level annotation model, which can effectively solve the problem of local image blur while ensuring the overall image denoising effect, and provides the possibility for denoising image edge information and restoring image details.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of image processing technology, and in particular to an image noise reduction method, device, equipment and storage medium. Background Art

[0002] With the popularization of various digital instruments and digital products, images, as the most commonly used information carrier in human activities, have gradually become the main way for people to obtain external information. However, in the process of image acquisition, transmission and storage, images are often interfered by various noises, which reduces the image quality. Therefore, it is necessary to perform noise reduction on the image to remove useless information in the signal and improve the image quality.

[0003] With the development of deep learning, image processing methods such as neural networks have been proposed in related technologies to reduce the noise of images. However, these noise reduction methods have poor noise reduction effects on locally blurred images. Summary of the invention

[0004] Embodiments of the present application provide an image noise reduction method, apparatus, device and storage medium to improve image noise reduction effect.

[0005] In a first aspect, an embodiment of the present application provides an image denoising method, comprising: inputting an image to be processed into an image denoising model for denoising processing, and obtaining an initial denoised image corresponding to the image to be processed output by the image denoising model; inputting the image to be processed into a labelling model for pixel labelling, and obtaining pixel information corresponding to the image to be processed output by the labelling model, wherein the pixel information is used to indicate the degree of blur of each pixel in the image to be processed; and obtaining a target denoised image corresponding to the image to be processed based on the pixel information and the initial denoised image.

[0006] It can be seen that in the embodiment of the present application, image denoising is achieved by combining an image denoising model with a pixel-level annotation model, which can effectively solve the problem of local image blur while ensuring the overall image denoising effect, and provides the possibility for denoising image edge information and restoring image details.

[0007] In a second aspect, an embodiment of the present application provides a training method for an image denoising model, comprising: obtaining a blurred sample image; inputting the blurred sample image into a labeling model for pixel labeling to obtain pixel information of the blurred sample image output by the labeling model, wherein the pixel information is used to indicate the degree of blur of each pixel in the blurred sample image; inputting the blurred sample image into an image denoising model for denoising to obtain an initial sample denoised image corresponding to the blurred sample image output by the image denoising model; and adjusting the model parameters of the image denoising model according to the pixel information and the initial sample denoised image.

[0008] It can be seen that in the embodiment of the present application, in the process of training the image denoising model, the image denoising model is trained using a pixel-level quality evaluation network (i.e., a labeling model), which can improve the accuracy of the image denoising model, so that the image denoising model can ensure the overall denoising effect of the image during the denoising process while also improving the image denoising model's processing effect on local blur.

[0009] In a third aspect, an embodiment of the present application provides an image denoising device, comprising: a denoising module, used to input the image to be processed into an image denoising model for denoising processing, and obtain an initial denoised image corresponding to the image to be processed output by the image denoising model; a labelling module, used to input the image to be processed into a labelling model for pixel labelling, and obtain pixel information corresponding to the image to be processed output by the labelling model, wherein the pixel information is used to indicate the degree of blur of each pixel in the image to be processed; and an acquisition module, used to obtain a target denoised image corresponding to the image to be processed based on the pixel information and the initial denoised image.

[0010] In a fourth aspect, an embodiment of the present application provides a training device for an image denoising model, comprising: an acquisition module for acquiring a blurred sample image; a labeling module for inputting the blurred sample image into the labeling model for pixel labeling, and obtaining pixel information of the blurred sample image output by the labeling model, wherein the pixel information is used to indicate the degree of blur of each pixel in the blurred sample image; a denoising module for inputting the blurred sample image into the image denoising model for denoising, and obtaining an initial sample denoised image corresponding to the blurred sample image output by the image denoising model; and a training module for adjusting the model parameters of the image denoising model according to the pixel information and the initial sample denoised image.

[0011] In a fifth aspect, an embodiment of the present application provides an electronic device, comprising: a memory and at least one processor, the memory being used to store program instructions, the processor being used to call the program instructions in the memory to execute an image denoising method such as the first aspect, and / or a training method for an image denoising model such as the second aspect.

[0012] In a sixth 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, an image denoising method as in the first aspect and / or a training method for an image denoising model as in the second aspect is implemented.

[0013] In a seventh aspect, an embodiment of the present application provides a computer program product, comprising: a computer program, which, when executed by a processor, implements the image denoising method of the first aspect, and / or the training method of the image denoising model of the second aspect.

[0014] The embodiment of the present application provides an image denoising method, apparatus, device and storage medium, wherein the image to be processed is input into an image denoising model for denoising, and an initial denoised image corresponding to the image to be processed output by the image denoising model is obtained; the image to be processed is input into an annotation model for pixel annotation, and pixel information corresponding to the image to be processed output by the annotation model is obtained, and the pixel information is used to indicate the blur degree of each pixel in the image to be processed; based on the pixel information and the initial denoised image, a target denoised image corresponding to the image to be processed is obtained. Image denoising is achieved by combining an image denoising model and a pixel-level annotation model, which can effectively solve the problem of local blurring of the image while ensuring the overall denoising effect of the image, and provides the possibility for denoising the edge information of the image and restoring the image details. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0016] Figure 1 A schematic diagram of a scene of an image noise reduction method provided in an embodiment of the present application;

[0017] Figure 2 A schematic diagram of a process for image noise reduction method provided in an embodiment of the present application;

[0018] Figure 3 A schematic diagram of the principle of the image noise reduction method provided in an embodiment of the present application;

[0019] Figure 4 A schematic diagram of the structure of a noise reduction network provided in an embodiment of the present application;

[0020] FIG5(a) is a flow chart of a training method for an image denoising model provided in an embodiment of the present application. Figure 1 ;

[0021] FIG5( b ) is a schematic diagram showing the principle of the training method of the image denoising model provided in the embodiment of the present application. Figure 1 ;

[0022] FIG6( a) is a flow chart of a training method for an image denoising model provided in an embodiment of the present application. Figure 2 ;

[0023] FIG6( b ) is a schematic diagram showing the principle of the training method of the image denoising model provided in the embodiment of the present application. Figure 2 ;

[0024] FIG. 7( a) is a flow chart of a training method for an image denoising model provided in an embodiment of the present application. Figure 3 ;

[0025] FIG. 7( b ) is a schematic diagram showing the principle of the training method of the image denoising model provided in the embodiment of the present application. Figure 3 ;

[0026] FIG8( a ) is a flow chart of a method for training an image denoising model provided in an embodiment of the present application. Figure 4 ;

[0027] FIG8( b ) is a schematic diagram showing the principle of the training method of the image denoising model provided in the embodiment of the present application. Figure 4 ;

[0028] Fig. 9 A schematic diagram of the structure of an image noise reduction device provided in an embodiment of the present application;

[0029] Fig.10 A schematic diagram of the structure of a training device for an image denoising model provided in an embodiment of the present application;

[0030] Fig.11 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0031] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0032] The terms "first" and "second" in the specification, claims and the above-mentioned drawings of the embodiments of the present application are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the implementation of the present application described herein can be implemented in an order other than the order shown or described. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0033] It should be understood that the term "and / or" used in this article is only a way to describe the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. " / " means the relationship of "or".

[0034] In order to improve image quality, it is often necessary to perform image noise reduction. With the development of deep learning, related technologies have proposed image processing methods such as neural networks to reduce image noise. However, among the existing noise reduction algorithms, some have good noise reduction effects but lose some image edge information, or are committed to studying the detection of image edge information and retaining image details. In general, the noise reduction effects of these noise reduction technologies cannot be guaranteed.

[0035] For example, for the DnCNN network, DnCNN uses batch normalization and residual learning to effectively remove uniform Gaussian noise and suppress noise within a certain noise level range. However, real noise is not uniform Gaussian noise, it is signal-dependent, related to each color channel, and is non-uniform and may vary with spatial position, making it difficult to guarantee the noise reduction effect.

[0036] Therefore, how to find a good balance between noise reduction and detail preservation has become a research focus in recent years. In view of this, the embodiments of the present application provide an image noise reduction method, device, equipment and storage medium. In the image noise reduction process, a dual-channel network of an image noise reduction model and a pixel-level quality evaluation network (i.e., an annotation model) is used to achieve image noise reduction. While ensuring the overall noise reduction effect of the image, the problem of local image blur can be effectively solved, which provides the possibility for noise reduction of image edge information and restoration of image details.

[0037] Next, the image denoising method is described in detail in conjunction with a specific embodiment. Figure 1 Schematic diagram of a scene of the image noise reduction method provided in the embodiment of the present application. Figure 1 As shown, the scenario includes: terminal equipment. The terminal equipment may also be referred to as user equipment (UE), mobile station (MS), mobile terminal, terminal, etc. In practical applications, terminal equipment includes, for example, desktop computers, notebooks, personal digital assistants (PDA), smart phones, tablet computers, vehicle-mounted equipment, wearable devices (such as smart watches and smart bracelets), smart home devices (such as smart display devices), etc.

[0038] Exemplarily, after the terminal device obtains the image to be processed (for example, the image to be processed is obtained by photographing the image with the terminal device, or the image to be processed is uploaded or sent to the terminal device by other means), the terminal device can perform noise reduction on the image to be processed by the image noise reduction method provided in the embodiment of the present application, and thereby obtain a noise-reduced image of the image to be processed.

[0039] In some optional embodiments, the scenario may also include a server. The server is a business point that provides data processing, database and other functions. The server may be an integrated server or a distributed server across multiple computers or computer data centers. The server may include hardware, software, or an embedded logic component or a combination of two or more such components for executing appropriate functions supported or implemented by the server. The server may be, for example, a blade server, a cloud server, etc., or may be a server group consisting of multiple servers.

[0040] The terminal device and the server can communicate with each other through a wired network or a wireless network. In the embodiment of the present application, the server can perform some functions of the above-mentioned terminal device. For example, the image to be processed can be uploaded to the server through the terminal device, and the server performs noise reduction processing on the image to be processed through the image noise reduction method provided in the embodiment of the present application, thereby obtaining a noise-reduced image of the image to be processed, and then the terminal device outputs the noise-reduced image.

[0041] It should be understood that Figure 1 This is only a schematic diagram of an application scenario provided by the embodiment of the present application. Figure 1 The types and number of devices included in the Figure 1 In the application scenario shown, a data storage device may also be included for storing business data. The data storage device may be an external memory or an internal memory integrated in a terminal device or a server.

[0042] The following specific embodiments are used to describe in detail the technical solutions of the embodiments of the present application and how the technical solutions of the present application solve the above-mentioned technical problems. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0043] Figure 2 It should be understood that the execution subject of the embodiment of the present application is the above-mentioned terminal device or server, such as Figure 2 As shown, the image denoising method comprises the following steps:

[0044] S201, inputting an image to be processed into an image denoising model for denoising, and obtaining an initial denoised image corresponding to the image to be processed output by the image denoising model.

[0045] It should be noted that the embodiments of the present application do not specifically limit the specific types of image denoising models. For example, the image denoising model in the embodiments of the present application may include but is not limited to one or more of the following: convolutional neural network denoising model (Denoising Convolutional Neural Network, DnCNN), FFDNet network model, CBDNet and mobilenetv model (for example, Mobilenetv2), etc., or the image denoising model may also be a combination or deformation of the above-mentioned network models.

[0046] It should be understood that for specific solutions of using the above-mentioned image denoising models to denoise the image to be processed, please refer to the relevant solutions in the prior art, and some examples are shown in the subsequent embodiments.

[0047] S202: Input the image to be processed into the annotation model for pixel annotation, and obtain pixel information corresponding to the image to be processed output by the annotation model.

[0048] The annotation model is obtained by training based on the blurred sample image and the clear sample image corresponding to the blurred sample image. The embodiment of the present application does not specifically limit the type of the annotation model. Exemplarily, the annotation model can be an image segmentation model, such as U2-Net.

[0049] In some embodiments, the pixel information is used to indicate the degree of blur of each pixel in the image to be processed. Exemplarily, taking the labeling model as U2-Net as an example, the pixel information includes the mask of the image to be processed. Specifically, the blurred pixels in the blurred sample image can be labeled as the first pixel information, and the clear pixels in the blurred sample image can be labeled as the second pixel information. In each round of training, the blurred sample image is used to train the labeling model to obtain the model labeling result output in the current round, and then the model labeling result and the actual labeling situation of the blurred sample image (i.e., the first pixel information or the second pixel information) are used to obtain the loss function corresponding to the current round, so as to adjust the parameters of the labeling model through the loss function, thereby obtaining a trained labeling model.

[0050] It should be noted that the embodiments of the present application do not specifically limit the specific sizes of the first pixel information and the second pixel information. By way of example, still taking the labeling model as U2-Net as an example, the mask value of the blurred pixel can be labeled as 1 (i.e., the first pixel information), and the mask value of the clear pixel can be labeled as 0.2 (i.e., the second pixel information).

[0051] It should be understood that there is no particular limitation on the execution order of the above steps S201 and S202.

[0052] S203: Obtain a target denoised image corresponding to the image to be processed based on the pixel information and the initial denoised image.

[0053] Still taking the example where the labeling model is U2-Net and the pixel information is the mask of the image to be processed, step S203 is specifically: multiplying the initial denoised image by the mask of the image to be processed to obtain a target denoised image corresponding to the image to be processed.

[0054] In an embodiment of the present application, image denoising is achieved by combining an image denoising model with a pixel annotation model. This can effectively solve the problem of local image blur while ensuring the overall image denoising effect, and provides the possibility for denoising image edge information and restoring image details.

[0055] In some optional embodiments, Figure 3 The schematic diagram is a principle diagram of the image noise reduction method provided in the embodiment of the present application. Figure 3 As shown, the image denoising model includes a denoising network. In the embodiment of the present application, the above step S201 specifically includes: performing denoising on the image to be processed through the denoising network to obtain an initial denoised image.

[0056] Next, take the image denoising model mobilenetv2 as an example, combined with Figure 4 The specific structure of the denoising network is described. It should be noted that when the image denoising model is mobilenetv2, the denoising network can be a network structure based on Bottleneck.

[0057] Figure 4 This is a schematic diagram of the structure of the noise reduction network provided in the embodiment of the present application. Figure 4 As shown, the denoising network includes: an expansion layer, a depth-separable convolution layer, and a projection layer. Specifically, when obtaining a target denoised image through the denoising network, the following steps are specifically included:

[0058] (1) Dimensionally expand the image to be processed through the expansion layer to obtain a first feature image;

[0059] (2) extracting features from the first feature image through a depthwise separable convolutional layer to obtain a second feature image;

[0060] (3) The second feature image is dimensionally compressed through a mapping layer to obtain an initial denoised image.

[0061] In some embodiments, the mapping layer can be a 1*1 network structure, which can map the second feature image from the high-dimensional feature to the low-dimensional space to obtain the initial denoised image. It should be noted that the design of mapping the high-dimensional space to the low-dimensional space using a 1*1 network structure can also be called a Bottleneck layer. Correspondingly, the expansion layer can be a 1*1 network structure, which is used to map the image to be processed from the low-dimensional space to the high-dimensional space to obtain the first feature image.

[0062] It should be noted that the expansion multiple of the expansion layer is not particularly limited in the embodiment of the present application. For example, the expansion multiple can be set to any value such as 3 times or 6 times. In the related art, the network structure of mobilenetv2 is shown in Table 1 below:

[0063] Table 1:

[0064]

[0065] In the embodiment of the present application, the network structure of mobilenetv2 is modified by at least one of the following:

[0066] (1) Remove the 7*7 average pooling layer (Avgpool);

[0067] (2) adjusting the step size s of each processing layer to a first value;

[0068] (3) Set the last 1*1 convolution layer (Conv2d) to a 1*1*3 convolution layer.

[0069] It should be noted that the embodiment of the present application does not limit the size of the first value, for example, the first value can be set to 1. Exemplarily, the network structure of the noise reduction network is shown in Table 2 below:

[0070] Table 2:

[0071]

[0072] Compared with the existing mobilenetv2 network structure, in the embodiment of the present application, by adjusting the step size S of each layer to the first value, the input and output sizes of each layer can no longer change. At the same time, by removing the pooling layer, the complexity of the image denoising model can be reduced.

[0073] Next, the training method of the image denoising model is described in detail in conjunction with a specific embodiment.

[0074] FIG5(a) is a flow chart of a training method for an image denoising model provided in an embodiment of the present application. Figure 1It should be understood that the execution subject of the embodiment of the present application may be the above-mentioned image noise reduction device, or may be other devices, and the embodiment of the present application is not particularly limited.

[0075] As shown in FIG5(a), the training method of the image denoising model includes the following steps:

[0076] S501: Obtain a blurred sample image.

[0077] S502: Input the blurred sample image into the annotation model for pixel annotation to obtain pixel information of the blurred sample image output by the annotation model.

[0078] The annotation model is a model pre-trained based on the blurred sample image, and the pixel information is used to indicate the blur degree of each pixel in the blurred sample image. It should be understood that the blurred sample images used to train the annotation model and the blurred sample images used to train the image denoising model can be the same data set, or can also be different data sets.

[0079] For a detailed method of training the annotation model by blurring the sample images, please refer to Figure 2 The embodiments shown will not be described in detail here.

[0080] S503: Input the blurred sample image into the image denoising model for denoising, and obtain an initial sample denoised image corresponding to the blurred sample image output by the image denoising model.

[0081] In some embodiments, the image denoising model includes: a denoising network, wherein the denoising network includes: an expansion layer, a depthwise separable convolution layer and a mapping layer. It should be understood that the specific structure of the denoising network and the implementation of each layer can be found in Figure 4 The embodiments shown will not be described in detail here.

[0082] In some optional embodiments, FIG5(b) is a schematic diagram of the principle of the training method of the image denoising model provided in the embodiment of the present application. Figure 1 As shown in FIG. 5( b ), the image denoising model in the embodiment of the present application further includes: an image augmentation network, which is used to perform augmentation processing on the blurred sample image to obtain an augmented image corresponding to the blurred sample image.

[0083] Furthermore, the augmented image and the blurred sample image are denoised by a denoising network to obtain an initial sample denoised image corresponding to the augmented image and an initial sample denoised image corresponding to the blurred sample image.

[0084] Specifically, when the denoising network obtains the initial sample denoised image, it specifically includes the following steps:

[0085] (1) The augmented image and the blurred sample image are dimensionally expanded through the expansion layer to obtain a third feature image output by the expansion layer;

[0086] (2) extracting features from the third feature image through a depthwise separable convolutional layer to obtain a fourth feature image output by the depthwise separable convolutional layer;

[0087] (3) The fourth feature image is dimensionally compressed through a mapping layer to obtain an initial sample denoised image output by the mapping layer.

[0088] It should be understood that the specific structure of the denoising network and the implementation of each layer can be found in Figure 4 The embodiments shown will not be described in detail here.

[0089] S504: Adjust model parameters of the image denoising model according to the pixel information and the initial sample denoised image.

[0090] It should be noted that the method of using pixel information and initial sample denoised images to train the image denoising model will be described in detail in subsequent embodiments.

[0091] In the embodiment of the present application, the image denoising model is trained using a pixel-level quality evaluation network (i.e., an annotation model), which can improve the accuracy of the image denoising model, so that the image denoising model can improve the processing effect of the image denoising model on local blur while ensuring the overall denoising effect of the image during the denoising process. In addition, by augmenting the image to be processed through the augmentation network, a series of random changes can be made to the image to be processed, and similar but different training samples can be generated, thereby expanding the scale of the training data set. At the same time, by randomly changing the training samples, the model's dependence on certain attributes can be reduced, and the generalization ability of the image denoising model can be improved.

[0092] In the first example of the present application, FIG6(a) is a flow chart of the training method of the image denoising model provided in the embodiment of the present application. Figure 2 FIG6( b ) is a schematic diagram showing the principle of the training method of the image denoising model provided in the embodiment of the present application. Figure 2 As shown in FIG. 6( a ) and FIG. 6( b ), the training method of the image denoising model provided in the embodiment of the present application specifically includes:

[0093] S601: Obtain a blurred sample image.

[0094] S602: Input the blurred sample image into the annotation model for pixel annotation to obtain pixel information of the blurred sample image output by the annotation model.

[0095] The annotation model is a model pre-trained based on the blurred sample image, and the pixel information is used to indicate the blur degree of each pixel in the blurred sample image;

[0096] S603: Input the blurred sample image into the image denoising model for denoising, and obtain an initial sample denoised image corresponding to the blurred sample image output by the image denoising model.

[0097] It should be noted that the principles and schemes of steps S601 to S603 are similar to those of steps S501 to S503 in the embodiment shown in FIG. 5( a ), and details may be referred to the above embodiment, which will not be described again here.

[0098] S604: Obtain a first loss value through a first loss function, pixel information, and the initial sample denoised image. The calculation formula of the first loss function is as follows:

[0099] Loss1=||HR-Mobile(LR)||1

[0100] Among them, Loss1 is the first loss value, LR is the image feature of each pixel in the initial sample denoising image, HR is the pixel information of each pixel in the blurred sample image. Exemplarily, taking the pixel information as a mask, if a pixel is a clear pixel, the HR value of the pixel is the second pixel information, if a pixel is a blurred pixel, the HR value of the pixel is the first pixel information.

[0101] S605: Adjust model parameters of the image denoising model based on the first loss value.

[0102] In an embodiment of the present application, the image denoising model is trained using a pixel-level quality evaluation network (i.e., a labeling model), which can improve the accuracy of the image denoising model, so that the image denoising model can improve the processing effect of the image denoising model on local blur while ensuring the overall denoising effect of the image during the denoising process.

[0103] In the second example of the present application, FIG. 7( a) is a flow chart of a training method for an image denoising model provided in an embodiment of the present application. Figure 3 FIG. 7( b) is a schematic diagram showing the principle of the training method of the image denoising model provided in the embodiment of the present application. Figure 3 As shown in FIG. 7( a ) and FIG. 7( b ), the training method of the image denoising model provided in the embodiment of the present application specifically includes the following steps:

[0104] S701, obtaining a blurred sample image;

[0105] S702: Input the blurred sample image into the annotation model for pixel annotation to obtain pixel information of the blurred sample image output by the annotation model.

[0106] The annotation model is a model pre-trained based on the blurred sample image, and the pixel information is used to indicate the blur degree of each pixel in the blurred sample image.

[0107] S703: Input the blurred sample image into the image denoising model for denoising, and obtain an initial sample denoised image corresponding to the blurred sample image output by the image denoising model.

[0108] S704: Obtain a target sample image through pixel information and the initial sample denoised image.

[0109] It should be noted that the principles and schemes of steps S601 to S603 are similar to those of steps S501 to S503 in the embodiment shown in FIG. 5 ; the scheme for obtaining the target sample image in step S704 is similar to that in FIG. Figure 2 The scheme of obtaining the target denoised image in step S203 in the illustrated embodiment is similar, and the details may refer to the above embodiment, which will not be described again here.

[0110] S705: Obtain a second loss value through a second loss function, pixel information, and target sample image.

[0111] Among them, the formula of the second loss function is as follows:

[0112] Loss2=||HR-(DR)||1

[0113] Among them, Loss2 is the second loss value, DR is the image feature of each pixel in the target sample denoising image, HR is the pixel information of each pixel in the blurred sample image. Exemplarily, taking the pixel information as a mask, if a pixel is a clear pixel, the HR value of the pixel is the second pixel information, if a pixel is a blurred pixel, the HR value of the pixel is the first pixel information.

[0114] S706: Adjust model parameters of the image denoising model according to the second loss value.

[0115] In a first implementation, the model parameters of the image denoising model may be adjusted based only on the second loss value to implement training of the image denoising model.

[0116] In the second implementation, the first loss value and the second loss value may be combined to simultaneously train the image denoising model. Specifically, the above step S706 specifically includes:

[0117] The model parameters of the image denoising model are adjusted through the first loss value and the second loss value.

[0118] It should be noted that the method for obtaining the first loss value can be found in Figure 6(a) to Figure 6(b) The embodiments shown will not be described in detail here.

[0119] In an embodiment of the present application, the image denoising model is trained using a pixel-level quality evaluation network (i.e., a labeling model), which can improve the accuracy of the image denoising model, so that the image denoising model can improve the processing effect of the image denoising model on local blur while ensuring the overall denoising effect of the image during the denoising process.

[0120] In addition, the image denoising model can be trained by using either the first loss value or the second loss value or a combination of the two, which can improve the accuracy of the image denoising model and further improve its image denoising effect.

[0121] In the third example of the present application, FIG8(a) is a flow chart of the training method of the image denoising model provided in the embodiment of the present application. Figure 4 FIG8( b ) is a schematic diagram showing the principle of the training method of the image denoising model provided in the embodiment of the present application. Figure 4 As shown in FIG8(a) and FIG8(b), the training method of the image denoising model provided in the embodiment of the present application specifically includes the following steps:

[0122] S801: Obtain a blurred sample image.

[0123] S802: Input the blurred sample image into the annotation model for pixel annotation to obtain pixel information of the blurred sample image output by the annotation model.

[0124] The annotation model is a model pre-trained based on the blurred sample image, and the pixel information is used to indicate the blur degree of each pixel in the blurred sample image.

[0125] S803: Input the blurred sample image into the image denoising model for denoising, and obtain an initial sample denoised image corresponding to the blurred sample image output by the image denoising model.

[0126] S804: Obtain a target sample image through pixel information and the initial sample denoised image.

[0127] It should be noted that the principles and schemes of steps S801 to S803 are similar to those of steps S501 to S503 in the embodiment shown in FIG. 5 ; the scheme for obtaining the target sample image in step S804 is similar to that in step S501 to S503 in the embodiment shown in FIG. 5 . Figure 2 The scheme of obtaining the target denoised image in step S203 in the illustrated embodiment is similar, and the details may refer to the above embodiment, which will not be described again here.

[0128] S805: Obtain a clear sample image corresponding to the blurred sample image.

[0129] S806. Obtain a third loss value through the perception model based on the third loss function, the target sample image and the clear sample image.

[0130] S807. Adjust the model parameters of the image denoising model according to the third loss value.

[0131] It should be noted that the embodiment of the present application does not specifically limit the type of the perception model. For example, the perception model may be a VGG16 network, etc., and accordingly, the third loss value may be a content loss function, which is used to indicate the similarity between the target sample image and the clear sample image.

[0132] Optionally, taking the perception model as a VGG16 network as an example, the third loss value may be the Euclidean distance between corresponding pixels of the target sample image and the clear sample image.

[0133] Exemplarily, as shown in FIG8( b ), the clear sample image and the target sample image corresponding to the blurred sample image are respectively input into the pre-trained perception model, and the Euclidean distance between the clear sample image and the target sample image obtained by the perception model through the k-th convolutional layer is represented by the j-th feature map output by the k-th convolutional layer of VGG16, thereby obtaining the third loss value:

[0134] Specifically, the formula of the third loss function is as follows:

[0135]

[0136] Among them, L VGG is the third loss value, G i (x) is the image feature of the target sample image, X i To clarify the image features of the sample image, N represents the total amount of feature maps output by the kth convolutional layer.

[0137] In the embodiment of the present application, by constraining the training process of the image denoising model through the perception model, the target denoised image obtained by the image denoising model can be made more realistic and the denoising effect can be improved.

[0138] In the first example, the above step S807 specifically includes:

[0139] (1) Obtaining a first loss value through a first loss function, pixel information, and an initial sample denoised image;

[0140] It should be noted that the scheme for obtaining the first loss value is similar to the scheme in the embodiments shown in FIG. 6( a ) and FIG. 6( b ). For details, please refer to the above embodiments and will not be described again here.

[0141] (2) Based on the first loss value and the third loss value, adjust the model parameters of the image denoising model.

[0142] Specifically, the target loss value can be determined according to the target loss function, and then the model parameters of the image denoising model can be adjusted using the target loss value.

[0143] For example, the target loss value Loss can be obtained by the following formula:

[0144] Loss = a1 × Loss1 + a3 × L VGG

[0145] Among them, a1 is the weight value of the first loss function, and a3 is the weight value of the third loss function. As for the specific values ​​of a1 and a3, the embodiment of the present application does not specifically limit them. For example, a1 and a3 can be any values ​​such as 0.5, 1, etc., and a1 and a3 can be the same value or different values.

[0146] In the second example, the above step S807 specifically includes:

[0147] (1) Obtaining a second loss value through a second loss function, pixel information, and a target sample image;

[0148] It should be noted that the scheme for obtaining the second loss value is similar to the scheme in the embodiments shown in FIG. 7( a ) and FIG. 7 ( b ). For details, please refer to the above embodiments and will not be described again here.

[0149] (2) Adjusting the model parameters of the image denoising model through the second loss value and the third loss value.

[0150] Specifically, the target loss value can be determined according to the target loss function, and then the model parameters of the image denoising model can be adjusted using the target loss value.

[0151] For example, the target loss value Loss can be obtained by the following formula:

[0152] Loss = a2 × Loss2 + a3 × L VGG

[0153] Among them, a2 is the weight value of the second loss function, and a3 is the weight value of the third loss function. As for the specific values ​​of a2 and a3, the embodiment of the present application does not make any special limitations, and a2 and a3 can be the same value or different values.

[0154] In the third example, the above step S807 specifically includes:

[0155] (1) Obtaining a first loss value through a first loss function, pixel information, and an initial sample denoised image;

[0156] (2) Obtaining a second loss value through a second loss function, pixel information, and a target sample image;

[0157] It should be noted that the scheme for obtaining the first loss value is similar to the scheme in the embodiments shown in Figures 6(a) and 6(b), and the scheme for obtaining the second loss value is similar to the scheme in the embodiments shown in Figures 7(a) and 7(b). For details, please refer to the above embodiments and will not be repeated here.

[0158] (3) Adjusting the model parameters of the image denoising model by using the first loss value, the second loss value and the third loss value. Specifically, the above step (3) specifically includes:

[0159] I. Based on the target loss function, the target loss value is obtained through the first loss value, the second loss value and the third loss value.

[0160] Among them, the target loss value Loss can be obtained by the following formula:

[0161] Loss=a1×Loss1+a2×Loss2+a3×L VGG

[0162] In a possible example, taking the value of a1 as 0.5, the value of a2 as 2, and the value of a3 as 0.5, then,

[0163] Loss=0.5Loss1+2Loss2+0.5L VGG

[0164] II. Adjust the model parameters of the image denoising model through the target loss value.

[0165] In the embodiment of the present application, the image denoising model is trained by the perception model, so that the target denoised image obtained by the image denoising model can be made more realistic, thereby improving the denoising effect of the image.

[0166] In addition, during the training process, the image denoising model is trained by any combination of the first loss value, the second loss value and the third loss value, so that the accuracy of the image denoising model can be improved, so that the image denoising model can meet the denoising requirements of various scenarios.

[0167] Fig. 9 This is a schematic diagram of the structure of the image noise reduction device provided in the embodiment of the present application. The image noise reduction device can be implemented by software and / or hardware. In practical applications, the image noise reduction device can be integrated in the terminal device or server as described above.

[0168] like Fig. 9 As shown, the image denoising device 900 includes: a denoising module 901, which is used to input the image to be processed into the image denoising model for denoising, and obtain an initial denoised image corresponding to the image to be processed output by the image denoising model;

[0169] The annotation module 902 is used to input the image to be processed into the annotation model for pixel annotation, and obtain pixel information corresponding to the image to be processed output by the annotation model, where the pixel information is used to indicate the blur degree of each pixel in the image to be processed;

[0170] The acquisition module 903 is used to obtain a target denoised image corresponding to the image to be processed based on the pixel information and the initial denoised image.

[0171] In some embodiments, the image denoising model includes: a denoising network, the denoising network includes: an expansion layer, a depth-separable convolution layer and a mapping layer, and the step length of each processing layer in the mapping layer is the first step length;

[0172] The denoising module 901 is specifically used for: performing dimension expansion on the image to be processed through the expansion layer to obtain a first feature image output by the expansion layer; performing feature extraction on the first feature image through the depthwise separable convolution layer to obtain a second feature image output by the depthwise separable convolution layer; performing dimension compression on the second feature image through the mapping layer to obtain an initial denoised image output by the mapping layer.

[0173] In some embodiments, the pixel information includes a mask of the image to be processed; the acquisition module 903 is specifically used to: multiply the initial denoised image by the mask of the image to be processed to obtain a target denoised image corresponding to the image to be processed.

[0174] It should be understood that the image denoising device 900 provided in the embodiment of the present application can be applied to the technical solution in the embodiment shown in the above-mentioned image denoising method, and its implementation principle and technical effect are similar, which will not be repeated here.

[0175] Fig.10 A schematic diagram of the structure of the training device of the image denoising model provided in the embodiment of the present application. It should be understood that the training device can be implemented by software and / or hardware. In practical applications, the speech recognition device can be integrated in the terminal device or server as above.

[0176] like Fig.10 As shown, the training device 1000 includes: an acquisition module 1001, used to acquire a blurred sample image;

[0177] The annotation module 1002 is used to input the blurred sample image into the annotation model for pixel annotation, and obtain pixel information of the blurred sample image output by the annotation model, where the pixel information is used to indicate the blur degree of each pixel in the blurred sample image;

[0178] The denoising module 1003 is used to input the blurred sample image into the image denoising model for denoising, and obtain an initial sample denoised image corresponding to the blurred sample image output by the image denoising model; the training module 1004 is used to adjust the model parameters of the image denoising model according to the pixel information and the initial sample denoised image.

[0179] In some embodiments, the image denoising model includes: an image augmentation network and a denoising network; the denoising module 1003 is specifically used to: perform augmentation processing on the blurred sample image through the image augmentation network to obtain an augmented image corresponding to the blurred sample image output by the augmentation network; perform denoising processing on the augmented image and the blurred sample image through the denoising network to obtain an initial sample denoised image output by the denoising network.

[0180] In some embodiments, the denoising network includes: an expansion layer, a depthwise separable convolution layer and a mapping layer, and the step size of each processing layer in the mapping layer is the first step size; the denoising module 1003 is specifically used to: dimensionally expand the augmented image and the blurred sample image through the expansion layer to obtain a third feature image output by the expansion layer; perform feature extraction on the third feature image through the depthwise separable convolution layer to obtain a fourth feature image output by the depthwise separable convolution layer; dimensionally compress the fourth feature image through the mapping layer to obtain an initial sample denoised image output by the mapping layer.

[0181] In some embodiments, the training module 1004 is specifically used to: obtain a first loss value through a first loss function, pixel information and an initial sample denoised image; and adjust model parameters of the image denoising model based on the first loss value.

[0182] In some embodiments, the training module 1004 is specifically used to: obtain a target sample image through pixel information and an initial sample denoised image; obtain a second loss value through a second loss function, as well as pixel information and a target sample image; and adjust the model parameters of the image denoising model through the first loss value and the second loss value.

[0183] In some embodiments, the training module 1004 is specifically used to: obtain a clear sample image corresponding to a blurred sample image; obtain a third loss value based on a third loss function, a target sample image and a clear sample image through a perceptual model, and the third loss value is used to indicate the similarity between the target sample image and the clear sample image; adjust the model parameters of the image denoising model through the first loss value, the second loss value and the third loss value.

[0184] In some embodiments, the training module 1004 is specifically used to: obtain a target loss value based on a target loss function through a first loss value, a second loss value, and a third loss value; and adjust model parameters of the image denoising model through the target loss value.

[0185] Fig.11 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. Fig.11 As shown, the electronic device 1100 includes: a processor 1101 , a memory 1102 , a communication interface 1103 and a system bus 1104 .

[0186] Among them, the memory 1102 and the communication interface 1103 are connected to the processor 1101 through the system bus 1104 and complete communication with each other. The memory 1102 is used to store program instructions, the communication interface 1103 is used to communicate with other devices, and the processor 1101 is used to call the program instructions in the memory to execute the image denoising method scheme of the above-mentioned method embodiment, and / or, execute the image denoising model training method scheme of the above-mentioned method embodiment.

[0187] Specifically, the processor 1101 may include one or more processing units, for example, the processor 1101 may be a central processing unit (CPU), a digital signal processor (DSP), an application specific integrated circuit (ASIC), etc. A general processor may be a microprocessor or the processor may be any conventional processor, etc. The steps of the method disclosed in the application may be directly embodied as being executed by a hardware processor, or may be executed by a combination of hardware and software modules in the processor.

[0188] The memory 1102 may be used to store program instructions. The memory 1102 may include a program storage area and a data storage area. Among them, the program storage area may store an operating system, an application required for at least one function (such as a sound playback function, etc.), etc. The data storage area may store data (such as audio data, etc.) created during the use of the electronic device 1100, etc. In addition, the memory 1102 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, a universal flash storage (UFS), etc. The processor 1101 executes various functional applications and data processing of the electronic device 1100 by running the program instructions stored in the memory 1102.

[0189] The communication interface 1103 can provide solutions for wireless communications including 2G / 3G / 4G / 110G, etc., applied to the electronic device 1100. The communication interface 1103 can receive electromagnetic waves through the antenna, filter, amplify, and process the received electromagnetic waves, and transmit them to the modulation and demodulation processor for demodulation. The communication interface 1103 can also amplify the signal modulated by the modulation and demodulation processor, and convert it into electromagnetic waves for radiation through the antenna.

[0190] In some embodiments, at least some functional modules of the communication interface 1103 may be disposed in the processor 1101 .

[0191] In some embodiments, at least some functional modules of the communication interface 1103 and at least some modules of the processor 1101 may be arranged in the same device.

[0192] The system bus 1104 may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The system bus 1104 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0193] It should be noted that the number of the memory 1102 and the processor 1101 is not limited in the embodiment of the present application, and they can be one or more. Fig.11 Taking one as an example for illustration; the memory 1102 and the processor 1101 can be connected by wire or wirelessly in various ways, such as through a bus connection.

[0194] In practical applications, the electronic device 1100 may be various forms of computers or mobile terminals, wherein the computer may be, for example, a laptop computer, a desktop computer, a workbench, a server, a blade server, a mainframe computer, etc.; the mobile terminal may be, for example, a personal digital assistant, a cellular phone, a smart phone, a wearable device, and other similar computing devices.

[0195] The electronic device of this embodiment can be used to execute the technical solution in the above method embodiment. Its implementation principle and technical effect are similar and will not be described in detail here.

[0196] An embodiment of the present application also provides a computer-readable storage medium having program instructions stored thereon. When the program instructions are executed, an image denoising method as in any of the above embodiments and / or a method for training an image denoising model are implemented.

[0197] An embodiment of the present application also provides a computer program product, including: a computer program, which, when executed by a processor, implements an image denoising method as in any of the above method embodiments, and / or a method for training an image denoising model.

[0198] In the above embodiments, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of modules is only a logical function division, and there may be other division methods in actual implementation, for example, multiple modules can be combined or integrated into another system, or some features can be ignored or not executed.

[0199] Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or modules, which can be electrical, mechanical or other forms. In addition, each functional module in each embodiment of the present application can be integrated into a processing unit, or each module can exist physically separately, or two or more modules can be integrated into one unit. The above-mentioned module-composed unit can be implemented in the form of hardware or in the form of hardware plus software functional units.

[0200] The above-mentioned integrated module implemented in the form of a software function module can be stored in a computer-readable storage medium. The above-mentioned software function module is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to perform some steps of the methods of various embodiments of the present application.

[0201] The above storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk, etc. The storage medium can be any available medium that can be accessed by a general or special-purpose computer.

[0202] Those skilled in the art can understand that all or part of the steps of implementing the above-mentioned method embodiments can be completed by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, the steps of the above-mentioned method embodiments are executed; and the aforementioned storage medium includes: ROM, RAM, disk or optical disk and other media that can store program codes.

Claims

1. An image denoising method, characterized in that: include: Inputting the image to be processed into an image denoising model for denoising, and obtaining an initial denoised image corresponding to the image to be processed output by the image denoising model; The image to be processed is input into the annotation model for pixel annotation, and the pixel information of the image to be processed output by the annotation model is obtained, and the pixel information is used to indicate the blur degree of each pixel in the image to be processed; wherein the image denoising model is obtained by adjusting the model parameters of the image denoising model based on a first loss value and a third loss value, and the first loss value is obtained by a first loss function, pixel information of a blurred sample image, and an initial sample denoised image corresponding to the blurred sample image; the third loss value is obtained by a perception model based on a third loss function, a target sample image, and a clear sample image; the target sample image is obtained by pixel information of the blurred sample image and the initial sample denoised image; the clear sample image is an image corresponding to the blurred sample image; the third loss value is used to indicate the similarity between the target sample image and the clear sample image; The pixel information is multiplied by the initial denoised image to obtain a target denoised image corresponding to the image to be processed.

2. The method according to claim 1, characterized in that The image denoising model comprises a denoising network, which comprises: an expansion layer, a depth-separable convolution layer and a mapping layer, wherein the step length of each processing layer in the mapping layer is the first step length; The step of inputting the image to be processed into the image denoising model for denoising to obtain an initial denoised image corresponding to the image to be processed output by the image denoising model comprises: Performing dimension expansion on the image to be processed by the expansion layer to obtain a first feature image output by the expansion layer; Performing feature extraction on the first feature image through the depthwise separable convolutional layer to obtain a second feature image output by the depthwise separable convolutional layer; The second feature image is dimensionally compressed by the mapping layer to obtain the initial denoised image output by the mapping layer.

3. The method according to claim 1 or 2, characterized in that: The pixel information includes a mask of the image to be processed; and the step of multiplying the pixel information with the initial denoised image to obtain a target denoised image corresponding to the image to be processed includes: The initial denoised image is multiplied by the mask of the image to be processed to obtain a target denoised image corresponding to the image to be processed.

4. A training method for an image denoising model, characterized in that: include: Get the blurred sample image; Inputting the blurred sample image into a labeling model for pixel labeling, and obtaining pixel information of the blurred sample image output by the labeling model, wherein the pixel information is used to indicate the blur degree of each pixel in the blurred sample image; Inputting the blurred sample image into an image denoising model for denoising, and obtaining an initial sample denoised image corresponding to the blurred sample image output by the image denoising model; Obtaining a first loss value through a first loss function, the pixel information and the initial sample denoised image; Obtain a target sample image through the pixel information and the initial sample denoised image; obtain a clear sample image corresponding to the blurred sample image; obtain a third loss value through a perception model based on a third loss function, the target sample image and the clear sample image, wherein the third loss value is used to indicate the similarity between the target sample image and the clear sample image; Based on the first loss value and the third loss value, the model parameters of the image denoising model are adjusted.

5. The method according to claim 4, characterized in that The image denoising model comprises: an image augmentation network and a denoising network; The step of inputting the blurred sample image into an image denoising model for denoising to obtain an initial sample denoised image corresponding to the blurred sample image output by the image denoising model comprises: Performing augmentation processing on the blurred sample image through the image augmentation network to obtain an augmented image corresponding to the blurred sample image output by the augmentation network; The augmented image and the blurred sample image are subjected to denoising processing by the denoising network to obtain the initial sample denoised image output by the denoising network.

6. The method according to claim 5, characterized in that The denoising network comprises an expansion layer, a depth-separable convolutional layer and a mapping layer, wherein the step length of each processing layer in the mapping layer is the first step length; The step of performing denoising on the augmented image and the blurred sample image through the denoising network to obtain the initial sample denoised image output by the denoising network includes: Performing dimension expansion on the augmented image and the blurred sample image through the expansion layer to obtain a third feature image output by the expansion layer; Extracting features of the third feature image through the depthwise separable convolutional layer to obtain a fourth feature image output by the depthwise separable convolutional layer; The fourth feature image is dimensionally compressed by the mapping layer to obtain the initial sample denoised image output by the mapping layer.

7. The method according to claim 6, characterized in that The adjusting the model parameters of the image denoising model based on the first loss value and the third loss value includes: Obtaining a target sample image through the pixel information and the initial sample denoised image; Obtaining a second loss value through a second loss function, the pixel information and the target sample image; The model parameters of the image denoising model are adjusted by using the first loss value, the second loss value and the third loss value.

8. The method according to claim 7, characterized in that The adjusting the model parameters of the image denoising model by using the first loss value, the second loss value and the third loss value includes: Based on the target loss function, obtaining a target loss value through the first loss value, the second loss value and the third loss value; The model parameters of the image denoising model are adjusted according to the target loss value.

9. An image noise reduction device, characterized in that: include: A denoising module, used for inputting the image to be processed into an image denoising model for denoising, and obtaining an initial denoised image corresponding to the image to be processed output by the image denoising model; A labeling module, used for inputting the image to be processed into a labeling model for pixel labeling, and obtaining pixel information corresponding to the image to be processed output by the labeling model, wherein the pixel information is used to indicate the blur degree of each pixel in the image to be processed; wherein the image denoising model is obtained by adjusting the model parameters of the image denoising model based on a first loss value and a third loss value, wherein the first loss value is obtained by a first loss function, pixel information of a blurred sample image, and an initial sample denoised image corresponding to the blurred sample image; the third loss value is obtained by a perception model based on a third loss function, a target sample image, and a clear sample image, wherein the target sample image is obtained by pixel information of the blurred sample image and the initial sample denoised image, and the clear sample image is an image corresponding to the blurred sample image; the third loss value is used to indicate the similarity between the target sample image and the clear sample image; The acquisition module is used to multiply the pixel information with the initial denoised image to obtain a target denoised image corresponding to the image to be processed.

10. A training device for an image denoising model, characterized in that: include: An acquisition module, used for acquiring a blurred sample image; A labeling module, used for inputting the blurred sample image into a labeling model for pixel labeling, and obtaining pixel information of the blurred sample image output by the labeling model, wherein the pixel information is used for indicating the blur degree of each pixel in the blurred sample image; A denoising module, used for inputting the blurred sample image into an image denoising model for denoising, and obtaining an initial sample denoised image corresponding to the blurred sample image output by the image denoising model; A training module is used to obtain a first loss value through a first loss function, the pixel information and the initial sample denoised image; obtain a target sample image through the pixel information and the initial sample denoised image; obtain a clear sample image corresponding to the blurred sample image; obtain a third loss value through a perception model based on a third loss function, the target sample image and the clear sample image, the third loss value being used to indicate the similarity between the target sample image and the clear sample image; and adjust the model parameters of the image denoising model based on the first loss value and the third loss value.

11. An electronic device, characterized in that: include: A memory and at least one processor, the memory is used to store program instructions, and the processor is used to call the program instructions in the memory to execute the image denoising method as described in any one of claims 1 to 3, and / or the image denoising model training method as described in any one of claims 4 to 8.

12. A computer-readable storage medium, characterized in that: A computer program is stored on the readable storage medium; when the computer program is executed, the image denoising method according to any one of claims 1 to 3 and / or the image denoising model training method according to any one of claims 4 to 8 are implemented.

13. A computer program product, characterized in that include: A computer program, which, when executed by a processor, implements the image denoising method according to any one of claims 1 to 3, and / or the image denoising model training method according to any one of claims 4 to 8.

Images