CN111369456A

CN111369456A - Image denoising method and device, electronic device and storage medium

Info

Publication number: CN111369456A
Application number: CN202010128779.0A
Authority: CN
Inventors: 徐璐; 张佳维
Original assignee: Shenzhen Sensetime Technology Co Ltd
Current assignee: Shenzhen Sensetime Technology Co Ltd
Priority date: 2020-02-28
Filing date: 2020-02-28
Publication date: 2020-07-03
Anticipated expiration: 2040-02-28
Also published as: CN111369456B

Abstract

The present disclosure relates to an image denoising method and apparatus, an electronic device, and a storage medium, the method including: carrying out image processing on a noisy image through a gradient information prediction network to obtain first prediction gradient information corresponding to the noisy image, wherein the first prediction gradient information is used for representing gradient change information of pixel points in a clear image corresponding to the noisy image; classifying the pixel points in the image with noise according to the first prediction gradient information to obtain a first classification result corresponding to each pixel point in the image with noise; and denoising the image with the noise according to the first classification result to obtain a denoised image corresponding to the image with the noise. The embodiment of the disclosure can obtain the denoised image with higher definition.

Description

Image denoising method and device, electronic device and storage medium

Technical Field

The present disclosure relates to the field of computer technologies, and in particular, to an image denoising method and apparatus, an electronic device, and a storage medium.

Background

An image is a visual information carrier through which people can obtain information. In general, the image generation and transmission processes are often interfered by noise, which increases the difficulty of image processing of the computer. The image denoising refers to performing image denoising on a noisy image to obtain a clear denoised image. The denoised image may also be subjected to other image processing operations, such as image segmentation, object recognition, etc. Therefore, image denoising has become an important research topic in the field of image processing and computer vision.

Disclosure of Invention

The disclosure provides an image denoising method and device, an electronic device and a storage medium.

According to an aspect of the present disclosure, there is provided an image denoising method, including: carrying out image processing on a noisy image through a gradient information prediction network to obtain first prediction gradient information corresponding to the noisy image, wherein the first prediction gradient information is used for representing gradient change information of pixel points in a clear image corresponding to the noisy image; classifying the pixel points in the image with noise according to the first prediction gradient information to obtain a first classification result corresponding to each pixel point in the image with noise; and denoising the image with the noise according to the first classification result to obtain a denoised image corresponding to the image with the noise.

The first prediction gradient information obtained by the gradient information prediction network prediction can be used for better reflecting the pixel gradient distribution of the clear image corresponding to the image with noise, so that a first classification result closer to the pixel gradient distribution in the clear image can be obtained based on the first prediction gradient information, the image with noise is denoised according to the first classification result, and the denoised image with higher definition can be obtained.

In a possible implementation manner, the classifying, according to the first prediction gradient information, the pixel points in the image with noise to obtain a first classification result corresponding to each pixel point in the image with noise includes: aiming at any pixel point in the image with noise, determining the gradient characteristic value of the pixel point according to the first prediction gradient information; determining a gradient category corresponding to the pixel point according to a gradient characteristic value interval in which the gradient characteristic value of the pixel point is positioned, wherein different gradient categories correspond to different gradient characteristic value intervals; and determining the first classification result according to the gradient classification corresponding to each pixel point.

The gradient characteristic value of each pixel point can be determined based on the first prediction gradient information, so that each pixel point can be classified by utilizing different gradient characteristic value intervals corresponding to different preset gradient categories, and a subsequent first classification result for classifying and denoising is effectively obtained.

In a possible implementation manner, the denoising the image with noise according to the first classification result to obtain a denoised image corresponding to the image with noise, includes: based on the first classification result, performing classification convolution denoising on different types of pixel points through a classification denoising network to obtain a denoised image corresponding to the image with noise.

The first classification result which is closer to the gradient distribution of pixels in the clear image is utilized to classify, convolve and denoise the pixels of different classes, so that the number of features in the convolution process can be reduced, and the denoising efficiency is effectively improved.

In one possible implementation, the classification denoising network includes a plurality of convolutional layers; based on the first classification result, performing classification convolution denoising on different types of pixel points through a classification denoising network to obtain a denoised image corresponding to the image with noise, including: based on the first classification result, performing classification convolution denoising on different types of pixel points in at least one convolution layer of the classification denoising network to obtain a denoised image corresponding to the image with noise.

According to the deep classification denoising network comprising the plurality of convolutional layers, classification convolution denoising is carried out on different types of pixel points in at least one convolutional layer, the number of features in the convolution process can be reduced, and then denoising efficiency is effectively improved.

In one possible implementation, the method further includes: performing image processing on a noisy sample image included in a training set through the gradient information prediction network to obtain second prediction gradient information corresponding to the noisy sample image, wherein the second prediction gradient information is used for representing gradient change information of pixel points in a clear image corresponding to the noisy sample image, and the training set further includes reference gradient information of the clear image corresponding to the noisy sample image; adjusting a network parameter of the gradient information prediction network based on a difference between the second prediction gradient information and the reference gradient information.

Through a network training process, a gradient information prediction network which meets the requirement can be obtained through training, and then in the actual denoising process of the image with noise, a gradient information prediction result which can better reflect the pixel gradient distribution of the clear image corresponding to the image with noise can be obtained through prediction according to the trained gradient information prediction network.

In a possible implementation manner, the training set further includes a clear image corresponding to the noisy sample image; the method further comprises the following steps: and determining the reference gradient information by carrying out structure tensor analysis on a clear image corresponding to the noisy sample image.

The structure tensor analysis is carried out on the clear image corresponding to the noise sample image in the training set in advance, and reference gradient information used for representing real gradient information of the clear image corresponding to the noise sample image can be determined, so that preparation can be made for subsequent training of the gradient information prediction network.

In a possible implementation manner, the training set further includes a clear image corresponding to the noisy sample image; the method further comprises the following steps: classifying pixel points in the noisy sample image according to the second prediction gradient information to obtain a second classification result corresponding to each pixel point in the noisy sample image; based on the second classification result, performing classification convolution denoising on different types of pixel points through the classification denoising network to obtain a denoised image corresponding to the noisy sample image; and adjusting the network parameters of the classified denoising network based on the difference between the denoised image corresponding to the noisy sample image and the clear image corresponding to the noisy sample image.

Through a network training process, a classification denoising network meeting requirements can be trained, and then in the actual denoising process of a noisy image, a first classification result of classifying pixel points can be obtained according to first prediction gradient information obtained based on a gradient information prediction network, and classification convolution denoising is carried out on different types of pixel points by utilizing the trained classification denoising network, so that the number of features in the convolution process can be reduced, and the denoising efficiency is effectively improved.

In one possible implementation, the gradient information includes at least one of: gradient direction, gradient strength and gradient consistency.

According to an aspect of the present disclosure, there is provided an image denoising apparatus including: the first gradient information prediction module is used for carrying out image processing on a noisy image through a gradient information prediction network to obtain first prediction gradient information corresponding to the noisy image, wherein the first prediction gradient information is used for representing gradient change information of pixel points in a clear image corresponding to the noisy image; the first classification module is used for classifying the pixel points in the image with noise according to the first prediction gradient information to obtain a first classification result corresponding to each pixel point in the image with noise; and the first denoising module is used for denoising the image with the noise according to the first classification result to obtain a denoised image corresponding to the image with the noise.

In one possible implementation manner, the first classification module includes: the first determining submodule is used for determining the gradient characteristic value of any pixel point in the image with noise according to the first prediction gradient information; the second determining submodule is used for determining the gradient category corresponding to the pixel point according to the gradient characteristic value interval in which the gradient characteristic value of the pixel point is positioned, wherein different gradient categories correspond to different gradient characteristic value intervals; and the third determining submodule is used for determining the first classification result according to the gradient classification corresponding to each pixel point.

In one possible implementation, the first denoising module is specifically configured to: based on the first classification result, performing classification convolution denoising on different types of pixel points through a classification denoising network to obtain a denoised image corresponding to the image with noise.

In one possible implementation, the classification denoising network includes a plurality of convolutional layers; the first denoising module is specifically configured to: based on the first classification result, performing classification convolution denoising on different types of pixel points in at least one convolution layer of the classification denoising network to obtain a denoised image corresponding to the image with noise.

In one possible implementation, the apparatus further includes: the second gradient information prediction module is used for performing image processing on a noisy sample image included in a training set through the gradient information prediction network to obtain second prediction gradient information corresponding to the noisy sample image, wherein the second prediction gradient information is used for representing gradient change information of pixel points in a clear image corresponding to the noisy sample image, and the training set further includes reference gradient information of the clear image corresponding to the noisy sample image; a first parameter adjustment module to adjust a network parameter of the gradient information prediction network based on a difference between the second predicted gradient information and the reference gradient information.

In a possible implementation manner, the training set further includes a clear image corresponding to the noisy sample image; the device further comprises: and the structure tensor analysis module is used for carrying out structure tensor analysis on the clear image corresponding to the noisy sample image to determine the reference gradient information.

In a possible implementation manner, the training set further includes a clear image corresponding to the noisy sample image; the device further comprises: the second classification module is used for classifying the pixel points in the noisy sample image according to the second prediction gradient information to obtain a second classification result corresponding to each pixel point in the noisy sample image; the second denoising module is used for carrying out classification convolution denoising on different types of pixel points through the classification denoising network based on the second classification result to obtain a denoised image corresponding to the sample image with noise; and the second network parameter adjusting module is used for adjusting the network parameters of the classified denoising network based on the difference between the denoised image corresponding to the noisy sample image and the clear image corresponding to the noisy sample image.

According to an aspect of the present disclosure, there is provided an electronic device including: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to invoke the memory-stored instructions to perform the above-described method.

According to an aspect of the present disclosure, there is provided a computer readable storage medium having stored thereon computer program instructions which, when executed by a processor, implement the above-described method.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure. Other features and aspects of the present disclosure will become apparent from the following detailed description of exemplary embodiments, which proceeds with reference to the accompanying drawings.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the disclosure.

FIG. 1 shows a flow diagram of an image denoising method according to an embodiment of the present disclosure;

FIG. 2 shows a block diagram of network training in accordance with an embodiment of the present disclosure;

FIG. 3 shows a block diagram of an image denoising apparatus according to an embodiment of the present disclosure;

FIG. 4 illustrates a block diagram of an electronic device of an embodiment of the disclosure;

fig. 5 shows a block diagram of an electronic device of an embodiment of the disclosure.

Detailed Description

Various exemplary embodiments, features and aspects of the present disclosure will be described in detail below with reference to the accompanying drawings. In the drawings, like reference numbers can indicate functionally identical or similar elements. While the various aspects of the embodiments are presented in drawings, the drawings are not necessarily drawn to scale unless specifically indicated.

The word "exemplary" is used exclusively herein to mean "serving as an example, embodiment, or illustration. Any embodiment described herein as "exemplary" is not necessarily to be construed as preferred or advantageous over other embodiments.

The term "and/or" herein is merely an association describing an associated object, meaning that three relationships may exist, e.g., a and/or B, may mean: a exists alone, A and B exist simultaneously, and B exists alone. In addition, the term "at least one" herein means any one of a plurality or any combination of at least two of a plurality, for example, including at least one of A, B, C, and may mean including any one or more elements selected from the group consisting of A, B and C.

Furthermore, in the following detailed description, numerous specific details are set forth in order to provide a better understanding of the present disclosure. It will be understood by those skilled in the art that the present disclosure may be practiced without some of these specific details. In some instances, methods, means, elements and circuits that are well known to those skilled in the art have not been described in detail so as not to obscure the present disclosure.

Fig. 1 shows a flowchart of an image denoising method according to an embodiment of the present disclosure. The image denoising method shown in fig. 1 may be performed by a terminal device or other processing device, wherein the terminal device may be a User Equipment (UE), a mobile device, a User terminal, a cellular phone, a cordless phone, a Personal Digital Assistant (PDA), a handheld device, a computing device, a vehicle-mounted device, a wearable device, or the like. The other processing devices may be servers or cloud servers, etc. In some possible implementations, the stereoscopic mesh deformation-based image processing method may be implemented by a processor calling computer-readable instructions stored in a memory. As shown in fig. 1, the method may include:

in step S11, the image processing method performs image processing on the noisy image through the gradient information prediction network to obtain first prediction gradient information corresponding to the noisy image, where the first prediction gradient information is used to indicate gradient change information of a pixel point in a sharp image corresponding to the noisy image.

In step S12, classifying the pixels in the noisy image according to the first prediction gradient information to obtain a first classification result corresponding to each pixel in the noisy image;

in step S13, denoising the noisy image according to the first classification result, to obtain a denoised image corresponding to the noisy image.

The image with noise is an image to be processed which needs to be subjected to image denoising, the clear image corresponding to the image with noise is a denoised ideal image corresponding to the image with noise, and the denoised image corresponding to the image with noise is a denoised actual image corresponding to the image with noise.

The image denoising method provided by the embodiment of the disclosure can be applied to image quality enhancement of photographing, image quality enhancement of security monitoring, and other denoising scenes needing denoising and image quality enhancement, and the disclosure is not particularly limited thereto.

In a possible implementation manner, classifying pixel points in the noisy image according to the first prediction gradient information to obtain a first classification result corresponding to each pixel point in the noisy image, includes: aiming at any pixel point in the image with noise, determining the gradient characteristic value of the pixel point according to the first prediction gradient information; determining a gradient category corresponding to the pixel point according to a gradient characteristic value interval in which the gradient characteristic value of the pixel point is positioned, wherein different gradient categories correspond to different gradient characteristic value intervals; and determining a first classification result according to the gradient classification corresponding to each pixel point.

In one possible implementation, the gradient information includes at least one of: gradient direction, gradient strength and gradient consistency. For any pixel point in the image, the gradient direction is used for representing the curve direction with the fastest change in the neighborhood of the pixel point, the gradient strength is used for representing the change value of the pixel point along the gradient direction, and the gradient consistency is used for representing the consistency of the curve direction in the neighborhood of the pixel point.

In one example, a value range (e.g., 0-pi) in a gradient direction is divided into M non-overlapping sections, a value range (e.g., 0-0.2) of gradient strength is divided into N non-overlapping sections, a value range (e.g., 0-1) of gradient consistency is divided into P non-overlapping sections, gradient characteristic value sections corresponding to a plurality of gradient categories can be obtained according to the section division of gradient information, when the gradient information only includes the gradient direction, gradient characteristic value sections corresponding to the M gradient categories can be obtained, when the gradient information only includes the gradient strength, gradient characteristic value sections corresponding to the N gradient categories can be obtained, when the gradient information only includes the gradient consistency, gradient characteristic value sections corresponding to the P gradient categories can be obtained, when the gradient information includes the gradient direction and the gradient strength, gradient characteristic value sections corresponding to the N gradient categories can be obtained, when the gradient information includes the gradient direction and the gradient consistency, gradient characteristic value sections corresponding to the M × N gradient categories can be obtained, when the gradient information includes the gradient direction and the gradient strength, the gradient characteristic value ranges corresponding to the gradient categories can be obtained, when the gradient information includes the gradient direction and the gradient consistency, the gradient characteristic value ranges include the gradient strength, when the gradient information includes the gradient consistency, the gradient characteristic value ranges of the gradient information includes the gradient classification of the M × P categories, the gradient classification can be obtained, when the gradient information, the gradient characteristic value ranges of the gradient information includes the gradient consistency, the gradient characteristic value ranges of the gradient information, the gradient classification of the gradient characteristic value ranges of the gradient classification of the gradient information, the gradient classification of.

For example, the range (0 to pi) of the gradient direction is divided into 2 non-overlapping sections: (0-pi/2) and (pi/2-pi), dividing the gradient intensity range (0-0.2) into 2 non-overlapping sections (0-0.1) and (0.1-0.2), dividing the gradient consistency range (0-1) into 2 non-overlapping sections (0-0.5) and (0.5-1), and obtaining gradient characteristic value sections corresponding to 8 gradient categories, wherein the gradient characteristic value sections corresponding to the first gradient category are the gradient direction (0-pi/2), the gradient intensity (0-0.1) and the gradient consistency (0-0.5), the gradient characteristic value sections corresponding to the second gradient category are the gradient direction (0-pi/2), the gradient intensity (0-0.1) and the gradient consistency (0.5-1), and the gradient characteristic value sections corresponding to the third gradient category are the gradient direction (0-pi/2), the gradient intensity (0-pi/2) and the gradient consistency (0.5-0.5), and the gradient characteristic value sections corresponding to the third gradient category are the gradient direction (0-pi/2), the gradient strength (0.1) and the gradient consistency (0.5-0.1) The gradient eigenvalue interval of (a) is: the gradient direction (0-pi/2) and the gradient strength (0.1-0.2) and the gradient consistency (0.5-1), the gradient characteristic value interval corresponding to the fifth gradient category is the gradient direction (pi/2-pi) and the gradient strength (0-0.1) and the gradient consistency (0-0.5), the gradient characteristic value interval corresponding to the sixth gradient category is the gradient direction (pi/2-pi) and the gradient strength (0-0.1) and the gradient consistency (0.5-1), the gradient characteristic value interval corresponding to the seventh gradient category is the gradient direction (pi/2-pi) and the gradient strength (0.1-0.2) and the gradient consistency (0-0.5), and the gradient characteristic value interval corresponding to the eighth gradient category is the gradient direction (pi/2-pi) and the gradient strength (0.1-0.2) and the gradient consistency (0.5-1).

The first prediction gradient information is a prediction result of gradient information of a clear image corresponding to the image with noise of the gradient information prediction network, and is used for representing gradient change information of pixel points in the clear image corresponding to the image with noise. The first prediction gradient information comprises a specific value of at least one of gradient direction, gradient strength and gradient consistency, the gradient characteristic value of each pixel point can be determined according to the first prediction gradient information, and then the gradient category corresponding to each pixel point is determined according to the gradient characteristic value interval in which the gradient characteristic value of each pixel point is located.

For example, the first predicted gradient information includes a gradient direction, a gradient strength, and a gradient consistency. For any pixel point in the image with noise, it can be determined that the gradient direction value corresponding to the pixel point is pi/3, the gradient intensity value is 0.15, and the gradient consistency value is 0.4 according to the first prediction gradient information, and then according to the gradient characteristic value interval corresponding to the 8 gradient categories, it can be determined that the gradient category corresponding to the pixel point is the third gradient category.

According to the gradient category corresponding to each pixel point in the image with noise, a first classification result for classifying each pixel point can be obtained.

In order to improve the processing efficiency, the noisy image may be divided into a plurality of image blocks in advance, wherein the size of the plurality of image blocks may be the same or different, and this disclosure does not specifically limit this.

The image processing is carried out on the image with noise through the gradient information prediction network, and the obtained first prediction gradient information can be first prediction gradient information corresponding to each image block in the image with noise. The gradient category corresponding to each image block is determined through the first prediction gradient information corresponding to each image block in the image with noise, and each image block is classified according to the gradient category corresponding to each image block, so that the classification of pixel points in the image with noise is realized, the first classification result is obtained, and the classification efficiency can be effectively improved.

In a possible implementation manner, denoising a noisy image according to a first classification result to obtain a denoised image corresponding to the noisy image, includes: based on the first classification result, performing classification convolution denoising on different types of pixel points through a classification denoising network to obtain a denoised image corresponding to the image with noise.

In one possible implementation, the classification denoising network includes a plurality of convolutional layers; based on the first classification result, carrying out classification convolution denoising on different types of pixel points through a classification denoising network to obtain a denoised image corresponding to the image with noise, and the method comprises the following steps: based on the first classification result, performing classification convolution denoising on different types of pixel points in at least one convolution layer of the classification denoising network to obtain a denoised image corresponding to the image with noise.

In an example, a classification denoising network can include a plurality of convolutional layers, for any of which classification convolutional denoising can be employed. For example, a classification denoising network includes two convolutional layers (a first convolutional layer and a second convolutional layer). Inputting a noisy image into a first convolution layer aiming at the first convolution layer, carrying out image processing on the noisy image in the first convolution layer through a gradient information prediction network to obtain first prediction gradient information of the noisy image, classifying pixel points in the noisy image according to the first prediction gradient information to obtain a first classification result, carrying out classification convolution denoising on the noisy image based on the first classification result to obtain a denoised feature map, and inputting the denoised feature map into a second convolution layer; and performing image processing on the denoised feature map through a gradient information prediction network aiming at the second convolution layer to obtain third prediction gradient information of the denoised feature map, classifying pixel points in the denoised feature map according to the third prediction gradient information to obtain a third classification result, classifying, convolving and denoising the denoised feature map based on the third classification result, and outputting a denoised image corresponding to the noisy image. The number of convolutional layers included in the classification denoising network can be determined according to actual conditions, the classification denoising network can perform classification convolutional denoising in all convolutional layers by adopting the method, and can also perform classification convolutional denoising in part of the convolutional layers by adopting the method, and the method is not particularly limited in the disclosure.

In the embodiment of the present disclosure, in order to implement the image denoising process by using the gradient information prediction network and the classification denoising network, network training may be performed on the gradient information prediction network and the classification denoising network in advance to obtain the gradient information prediction network and the classification denoising network meeting the requirements. The following describes the network training process for the gradient information prediction network and the classification denoising network in detail.

In one possible implementation, the method further includes: performing image processing on the noisy sample image included in the training set through a gradient information prediction network to obtain second predicted gradient information corresponding to the noisy sample image, wherein the second predicted gradient information is used for representing gradient change information of pixel points in a clear image corresponding to the noisy sample image, and reference gradient information of the clear image corresponding to the noisy sample image is also included in the training set; adjusting network parameters of the gradient information prediction network based on a difference between the second prediction gradient information and the reference gradient information.

In the embodiment of the disclosure, a training set may be created in advance, and the training set may include reference gradient information of a noisy sample image and a clear image corresponding to the noisy sample image, so that a gradient information prediction network may be trained by the training set. The gradient information prediction network may be a neural network, or may be other network models that can implement gradient information prediction, which is not specifically limited in this disclosure. Similarly, in order to improve the processing efficiency, the noisy sample image may be divided into a plurality of image blocks in advance, where the size of the plurality of image blocks may be the same or different, and this disclosure does not specifically limit this.

In a possible implementation manner, the training set further includes a clear image corresponding to the noisy sample image, and the method further includes: and determining reference gradient information by carrying out structure tensor analysis on the clear image corresponding to the noisy sample image.

Before the gradient information prediction network is trained by using the training set, structure tensor analysis is carried out on a clear image corresponding to a noisy sample image in the training set in advance, reference gradient information used for representing real gradient information of the clear image corresponding to the noisy sample image can be determined, and therefore preparation can be made for subsequent training of the gradient information prediction network. The reference gradient information is used for representing gradient change information of pixel points in a clear image corresponding to the noisy sample image. Since the noisy sample image is divided into a plurality of image blocks, the clear image corresponding to the noisy sample image is also correspondingly divided into a plurality of image blocks. The reference gradient information is a specific value of at least one of the gradient direction, the gradient strength and the gradient consistency of each image block in the clear image corresponding to the noisy sample image.

In an example, for any image block (e.g., 7 × 7 image block) in the clear image corresponding to the noisy sample image, gradient values of each pixel point in the image block in the clear image corresponding to the noisy sample image in the x direction and the y direction are determined, so as to obtain a gradient matrix G, where G is used to obtain a gradient matrix G^TG yields a feature matrix. Performing characteristic analysis on the characteristic matrix to obtain a characteristic value lambda₁And λ₂And a characteristic value λ₁Corresponding feature vector phi₁And a characteristic value lambda₂Corresponding feature vector phi₂. Wherein λ is₁>λ₂。

The specific value theta of the gradient direction of the image block in the clear image corresponding to the image with the noise sample can be determined according to the following formula (I)_G：

θ_G＝arctan(φ_1,y,φ_1,x) (one) of the two types of the raw materials,

wherein phi is_1,xIs a feature vector phi₁Projection in the x direction, phi_1,yIs a feature vector phi₁Projection in the y-direction.

The specific value lambda of the gradient intensity of the image block in the clear image corresponding to the noisy sample image can be determined according to the following formula (II)_G：

λ_G＝λ₁And (II) performing secondary treatment.

According to the following formula (III), the specific value mu of the gradient consistency of the image block in the clear image corresponding to the image with the noise sample can be determined_G:

According to the specific value theta of the gradient direction of the image block in the clear image corresponding to the image with the noise sample_GSpecific value of gradient strength lambda_GSpecific value mu consistent with gradient_GReference gradient information corresponding to the image block in the sharp image corresponding to the noisy sample image may be determined.

And training the gradient information prediction network by using the reference gradient information of the noisy sample image and the clear image corresponding to the noisy sample image in the training set.

Fig. 2 illustrates a block diagram of network training in accordance with an embodiment of the disclosure. As shown in fig. 2, the noisy sample image in the training set is input to the gradient information prediction network, and after the image processing is performed on the noisy sample image by the gradient information prediction network, second prediction gradient information corresponding to the noisy sample image is output. The second prediction gradient information is a prediction result of the gradient information of the clear image corresponding to the noisy sample image by the gradient information prediction network, and the second prediction gradient information is a specific value of at least one of the gradient direction, the gradient strength and the gradient consistency of each image block in the noisy sample image. And determining the difference between the second prediction gradient information and the reference gradient information of the clear image corresponding to the noisy sample image included in the training set according to the gradient information loss function, namely determining the prediction loss of the gradient information prediction network, and further adjusting the network parameters of the gradient information prediction network according to the backward propagation of the prediction loss. The gradient information loss function may be a squared error loss function, or may also be other loss functions, which is not specifically limited in this disclosure.

Through the network training process, a gradient information prediction network which meets the requirement can be obtained through training, in the actual denoising process of the image with noise, a gradient information prediction result which can better reflect the pixel gradient distribution of the clear image corresponding to the image with noise can be obtained through prediction according to the trained gradient information prediction network, a pixel classification result which is closer to the pixel gradient distribution in the clear image corresponding to the image with noise can be obtained according to the gradient information prediction result, the image with noise can be denoised according to the pixel classification result, and the denoised image with higher definition can be obtained.

And training the classification denoising network by using the gradient information prediction network and the clear images which are included in the training set and correspond to the noisy sample images. The classification denoising network may be a neural network, or may be other network models that can implement classification denoising, and this disclosure does not specifically limit this.

In a possible implementation manner, the training set further includes a clear image corresponding to the noisy sample image; the method further comprises the following steps: classifying the pixel points in the noisy sample image according to the second prediction gradient information to obtain a second classification result corresponding to each pixel point in the noisy sample image; based on a second classification result, performing classification convolution denoising on different types of pixel points through a classification denoising network to obtain a denoised image corresponding to a sample image with noise; and adjusting network parameters of the classified denoising network based on the difference between the denoised image corresponding to the noise sample image and the clear image corresponding to the noise sample image.

Still taking the above fig. 2 as an example, as shown in fig. 2, according to the second prediction gradient information of the clear image corresponding to the noisy sample image output by the gradient information prediction network, classifying the plurality of image blocks in the noisy sample image to realize classification of the pixel points in the noisy sample image, and obtaining a second classification result. And inputting the noisy sample image and the second classification result into a classification denoising network, performing classification convolution denoising on different types of pixel points in the noisy sample image by the classification denoising network based on the second classification result (for example, performing convolution denoising operation by adopting different convolution weight kernels aiming at the different types of pixel points), and outputting a denoised image corresponding to the noisy sample image. And determining the difference between the denoised image corresponding to the sample image with noise and the clear image corresponding to the sample image with noise according to the denoising loss function, namely determining the denoising loss of the classified denoising network, and further adjusting the network parameters of the classified denoising network according to the denoising loss back propagation. The denoising loss function may be a square error loss function, or may also be other loss functions, which is not specifically limited in this disclosure.

In one example, the convolution kernel sizes corresponding to the different classes of pixel points are the same but take different values to achieve that the different classes of pixel points correspond to different convolution weights. For example, for a certain class of pixel points (x, y) in the noisy sample image, a convolution denoising operation is performed by the following formula (four):

wherein Q' (x, y, c) is the output characteristic of the pixel point (x, y) in the noisy sample image after denoising on the output channel c,

k is the size of the convolution kernel corresponding to the pixel point (x, y), m and n are two variables in the convolution denoising process, C_inThe total number of input channels corresponding to the pixel (x, y), c ' is a certain input channel corresponding to the pixel (x, y), and Q (x, y, c ') is the input characteristic of the pixel (x, y) in the noisy sample image on the input channel c '.

In one example, the convolution kernel sizes corresponding to different classes of pixel points are different to achieve different convolution weights corresponding to different classes of pixel points. The pixel points of different classes corresponding to different convolution weights may also be implemented in other manners, which is not specifically limited by this disclosure.

Through the network training process, a classification denoising network meeting requirements can be trained, and then in the actual denoising process of the image with noise, classification convolution denoising is carried out on different types of pixel points in the image with noise by using the trained classification denoising network according to a classification result which is obtained based on a gradient information prediction network and is closer to the pixel gradient distribution in a clear image corresponding to the image with noise, so that the characteristic quantity in the convolution process can be reduced, and the denoising efficiency is effectively improved.

The first prediction gradient information obtained by the gradient information prediction network prediction can be used for better reflecting the pixel gradient distribution of the clear image corresponding to the image with noise, so that a first classification result closer to the pixel gradient distribution in the clear image can be obtained based on the first prediction gradient information, and different types of pixel points in the image with noise are classified, convolved and denoised by using the first classification result closer to the pixel gradient distribution in the clear image, so that the number of features in the convolution process can be reduced, the denoising efficiency is effectively improved, and a denoised image with higher definition is obtained.

It is understood that the above-mentioned method embodiments of the present disclosure can be combined with each other to form a combined embodiment without departing from the logic of the principle, which is limited by the space, and the detailed description of the present disclosure is omitted. Those skilled in the art will appreciate that in the above methods of the specific embodiments, the specific order of execution of the steps should be determined by their function and possibly their inherent logic.

In addition, the present disclosure also provides an image denoising device, an electronic device, a computer-readable storage medium, and a program, which can be used to implement any image denoising method provided by the present disclosure, and the corresponding technical solutions and descriptions thereof will be referred to the corresponding records of the method section, and will not be described again.

Fig. 3 illustrates a block diagram of an image denoising apparatus according to an embodiment of the present disclosure. As shown in fig. 3, the apparatus 30 includes:

the first gradient information prediction module 31 is configured to perform image processing on the image with noise through a gradient information prediction network to obtain first prediction gradient information corresponding to the image with noise, where the first prediction gradient information is used to indicate gradient change information of a pixel point in a sharp image corresponding to the image with noise;

the first classification module 32 is configured to classify the pixel points in the image with noise according to the first prediction gradient information to obtain a first classification result corresponding to each pixel point in the image with noise;

and the first denoising module 33 is configured to denoise the noisy image according to the first classification result to obtain a denoised image corresponding to the noisy image.

In one possible implementation, the first classification module 32 includes:

the first determining submodule is used for determining the gradient characteristic value of any pixel point in the image with noise according to the first prediction gradient information;

the second determining submodule is used for determining the gradient category corresponding to the pixel point according to the gradient characteristic value interval in which the gradient characteristic value of the pixel point is positioned, wherein different gradient categories correspond to different gradient characteristic value intervals;

and the third determining submodule is used for determining a first classification result according to the gradient classification corresponding to each pixel point.

In one possible implementation, the first denoising module 33 is specifically configured to:

based on the first classification result, performing classification convolution denoising on different types of pixel points through a classification denoising network to obtain a denoised image corresponding to the image with noise.

In one possible implementation, the classification denoising network includes a plurality of convolutional layers;

the first denoising module 33 is specifically configured to:

based on the first classification result, performing classification convolution denoising on different types of pixel points in at least one convolution layer of the classification denoising network to obtain a denoised image corresponding to the image with noise.

In one possible implementation, the apparatus 30 further includes:

the second gradient information prediction module is used for carrying out image processing on the noisy sample image included in the training set through a gradient information prediction network to obtain second prediction gradient information corresponding to the noisy sample image, wherein the second prediction gradient information is used for representing gradient change information of pixel points in a clear image corresponding to the noisy sample image, and the training set further comprises reference gradient information of the clear image corresponding to the noisy sample image;

and the first parameter adjusting module is used for adjusting the network parameters of the gradient information prediction network based on the difference between the second prediction gradient information and the reference gradient information.

In a possible implementation manner, the training set further includes a clear image corresponding to the noisy sample image;

the apparatus 30 further comprises:

and the structure tensor analysis module is used for carrying out structure tensor analysis on the clear image corresponding to the noisy sample image to determine reference gradient information.

the apparatus 30 further comprises:

the second classification module is used for classifying the pixel points in the noisy sample image according to the second prediction gradient information to obtain a second classification result corresponding to each pixel point in the noisy sample image;

the second denoising module is used for carrying out classification convolution denoising on the pixel points of different classes through a classification denoising network based on a second classification result to obtain a denoised image corresponding to the sample image with noise;

and the second network parameter adjusting module is used for adjusting the network parameters of the classified denoising network based on the difference between the denoised image corresponding to the noise sample image and the clear image corresponding to the noise sample image.

In some embodiments, functions of or modules included in the apparatus provided in the embodiments of the present disclosure may be used to execute the method described in the above method embodiments, and specific implementation thereof may refer to the description of the above method embodiments, and for brevity, will not be described again here.

Embodiments of the present disclosure also provide a computer-readable storage medium having stored thereon computer program instructions, which when executed by a processor, implement the above-mentioned method. The computer readable storage medium may be a non-volatile computer readable storage medium.

An embodiment of the present disclosure further provides an electronic device, including: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to invoke the memory-stored instructions to perform the above-described method.

The embodiments of the present disclosure also provide a computer program product, which includes computer readable code, and when the computer readable code is run on a device, a processor in the device executes instructions for implementing the image denoising method provided in any one of the above embodiments.

The embodiments of the present disclosure also provide another computer program product for storing computer readable instructions, which when executed, cause a computer to perform the operations of the image denoising method provided in any of the above embodiments.

The electronic device may be provided as a terminal, server, or other form of device.

Fig. 4 shows a block diagram of an electronic device in accordance with an embodiment of the disclosure. As shown in fig. 4, the electronic device 800 may be a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, or the like terminal.

Referring to fig. 4, electronic device 800 may include one or more of the following components: processing component 802, memory 804, power component 806, multimedia component 808, audio component 810, input/output (I/O) interface 812, sensor component 814, and communication component 816.

The processing component 802 generally controls overall operation of the electronic device 800, such as operations associated with display, telephone calls, data communications, camera operations, and recording operations. The processing components 802 may include one or more processors 820 to execute instructions to perform all or a portion of the steps of the methods described above. Further, the processing component 802 can include one or more modules that facilitate interaction between the processing component 802 and other components. For example, the processing component 802 can include a multimedia module to facilitate interaction between the multimedia component 808 and the processing component 802.

The memory 804 is configured to store various types of data to support operations at the electronic device 800. Examples of such data include instructions for any application or method operating on the electronic device 800, contact data, phonebook data, messages, pictures, videos, and so forth. The memory 804 may be implemented by any type or combination of volatile or non-volatile memory devices 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 or optical disks.

The power supply component 806 provides power to the various components of the electronic device 800. The power components 806 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power for the electronic device 800.

The multimedia component 808 includes a screen that provides an output interface between the electronic device 800 and a user. In some embodiments, the screen may include a Liquid Crystal Display (LCD) and a Touch Panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive an input signal from a user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor may not only sense the boundary of a touch or slide action, but also detect the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 808 includes a front facing camera and/or a rear facing camera. The front camera and/or the rear camera may receive external multimedia data when the electronic device 800 is in an operation mode, such as a shooting mode or a video mode. Each front camera and rear camera may be a fixed optical lens system or have a focal length and optical zoom capability.

The audio component 810 is configured to output and/or input audio signals. For example, the audio component 810 includes a Microphone (MIC) configured to receive external audio signals when the electronic device 800 is in an operational mode, such as a call mode, a recording mode, and a voice recognition mode. The received audio signals may further be stored in the memory 804 or transmitted via the communication component 816. In some embodiments, audio component 810 also includes a speaker for outputting audio signals.

The I/O interface 812 provides an interface between the processing component 802 and peripheral interface modules, which may be keyboards, click wheels, buttons, etc. These buttons may include, but are not limited to: a home button, a volume button, a start button, and a lock button.

The sensor assembly 814 includes one or more sensors for providing various aspects of state assessment for the electronic device 800. For example, the sensor assembly 814 may detect an open/closed state of the electronic device 800, the relative positioning of components, such as a display and keypad of the electronic device 800, the sensor assembly 814 may also detect a change in the position of the electronic device 800 or a component of the electronic device 800, the presence or absence of user contact with the electronic device 800, orientation or acceleration/deceleration of the electronic device 800, and a change in the temperature of the electronic device 800. Sensor assembly 814 may include a proximity sensor configured to detect the presence of a nearby object without any physical contact. The sensor assembly 814 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 814 may also include an acceleration sensor, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.

The communication component 816 is configured to facilitate wired or wireless communication between the electronic device 800 and other devices. The electronic device 800 may access a wireless network based on a communication standard, such as WiFi, 2G or 3G, or a combination thereof. In an exemplary embodiment, the communication component 816 receives a broadcast signal or broadcast related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 816 further includes a Near Field Communication (NFC) module to facilitate short-range communications. For example, the NFC module may be implemented based on Radio Frequency Identification (RFID) technology, infrared data association (IrDA) technology, Ultra Wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

In an exemplary embodiment, the electronic device 800 may be implemented by one or more Application Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), Digital Signal Processing Devices (DSPDs), Programmable Logic Devices (PLDs), Field Programmable Gate Arrays (FPGAs), controllers, micro-controllers, microprocessors or other electronic components for performing the above-described methods.

In an exemplary embodiment, a non-transitory computer-readable storage medium, such as the memory 804, is also provided that includes computer program instructions executable by the processor 820 of the electronic device 800 to perform the above-described methods.

Fig. 5 shows a block diagram of an electronic device in accordance with an embodiment of the disclosure. As shown in fig. 5, the electronic device 1900 may be provided as a server. Referring to fig. 5, electronic device 1900 includes a processing component 1922 further including one or more processors and memory resources, represented by memory 1932, for storing instructions, e.g., applications, executable by processing component 1922. The application programs stored in memory 1932 may include one or more modules that each correspond to a set of instructions. Further, the processing component 1922 is configured to execute instructions to perform the above-described method.

The electronic device 1900 may also include a power component 1926 configured to perform power management of the electronic device 1900, a wired or wireless network interface 1950 configured to connect the electronic device 1900 to a network, and an input/output (I/O) interface 1958. The electronic device 1900 may operate based on an operating system, such as Windows Server, stored in memory 1932^TM，Mac OS X^TM，Unix^TM,Linux^TM，FreeBSD^TMOr the like.

In an exemplary embodiment, a non-transitory computer readable storage medium, such as the memory 1932, is also provided that includes computer program instructions executable by the processing component 1922 of the electronic device 1900 to perform the above-described methods.

The present disclosure may be systems, methods, and/or computer program products. The computer program product may include a computer-readable storage medium having computer-readable program instructions embodied thereon for causing a processor to implement various aspects of the present disclosure.

The computer readable storage medium may be a tangible device that can hold and store the instructions for use by the instruction execution device. The computer readable storage medium may be, for example, but not limited to, an electronic memory device, a magnetic memory device, an optical memory device, an electromagnetic memory device, a semiconductor memory device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium would include the following: a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a Static Random Access Memory (SRAM), a portable compact disc read-only memory (CD-ROM), a Digital Versatile Disc (DVD), a memory stick, a floppy disk, a mechanical coding device, such as punch cards or in-groove projection structures having instructions stored thereon, and any suitable combination of the foregoing. Computer-readable storage media as used herein is not to be construed as transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through a waveguide or other transmission medium (e.g., optical pulses through a fiber optic cable), or electrical signals transmitted through electrical wires.

The computer-readable program instructions described herein may be downloaded from a computer-readable storage medium to a respective computing/processing device, or to an external computer or external storage device via a network, such as the internet, a local area network, a wide area network, and/or a wireless network. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers and/or edge servers. The network adapter card or network interface in each computing/processing device receives computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in a computer-readable storage medium in the respective computing/processing device.

The computer program instructions for carrying out operations of the present disclosure may be assembler instructions, Instruction Set Architecture (ISA) instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source or object code written in any combination of one or more programming languages, including an object oriented programming language such as Smalltalk, C + + or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The computer-readable program instructions may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet service provider). In some embodiments, the electronic circuitry that can execute the computer-readable program instructions implements aspects of the present disclosure by utilizing the state information of the computer-readable program instructions to personalize the electronic circuitry, such as a programmable logic circuit, a Field Programmable Gate Array (FPGA), or a Programmable Logic Array (PLA).

Various aspects of the present disclosure are described herein with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the disclosure. It will be understood that each block of the flowchart illustrations and/or block diagrams, and combinations of blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer-readable program instructions.

These computer-readable program instructions may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions/acts specified in the flowchart and/or block diagram block or blocks. These computer-readable program instructions may also be stored in a computer-readable storage medium that can direct a computer, programmable data processing apparatus, and/or other devices to function in a particular manner, such that the computer-readable medium storing the instructions comprises an article of manufacture including instructions which implement the function/act specified in the flowchart and/or block diagram block or blocks.

The computer readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other devices to cause a series of operational steps to be performed on the computer, other programmable apparatus or other devices to produce a computer implemented process such that the instructions which execute on the computer, other programmable apparatus or other devices implement the functions/acts specified in the flowchart and/or block diagram block or blocks.

The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of instructions, which comprises one or more executable instructions for implementing the specified logical function(s). In some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

The computer program product may be embodied in hardware, software or a combination thereof. In an alternative embodiment, the computer program product is embodied in a computer storage medium, and in another alternative embodiment, the computer program product is embodied in a Software product, such as a Software Development Kit (SDK), or the like.

Having described embodiments of the present disclosure, the foregoing description is intended to be exemplary, not exhaustive, and not limited to the disclosed embodiments. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen in order to best explain the principles of the embodiments, the practical application, or improvements made to the technology in the marketplace, or to enable others of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. An image denoising method, comprising:

carrying out image processing on a noisy image through a gradient information prediction network to obtain first prediction gradient information corresponding to the noisy image, wherein the first prediction gradient information is used for representing gradient change information of pixel points in a clear image corresponding to the noisy image;

classifying the pixel points in the image with noise according to the first prediction gradient information to obtain a first classification result corresponding to each pixel point in the image with noise;

and denoising the image with the noise according to the first classification result to obtain a denoised image corresponding to the image with the noise.

2. The method of claim 1, wherein the classifying the pixels in the noisy image according to the first prediction gradient information to obtain a first classification result corresponding to each pixel in the noisy image comprises:

aiming at any pixel point in the image with noise, determining the gradient characteristic value of the pixel point according to the first prediction gradient information;

determining a gradient category corresponding to the pixel point according to a gradient characteristic value interval in which the gradient characteristic value of the pixel point is positioned, wherein different gradient categories correspond to different gradient characteristic value intervals;

and determining the first classification result according to the gradient classification corresponding to each pixel point.

3. The method according to claim 1 or 2, wherein the denoising the noisy image according to the first classification result to obtain a denoised image corresponding to the noisy image comprises:

4. The method of claim 3, wherein the classification denoising network comprises a plurality of convolutional layers;

based on the first classification result, performing classification convolution denoising on different types of pixel points through a classification denoising network to obtain a denoised image corresponding to the image with noise, including:

5. The method according to any one of claims 1 to 4, further comprising:

performing image processing on a noisy sample image included in a training set through the gradient information prediction network to obtain second prediction gradient information corresponding to the noisy sample image, wherein the second prediction gradient information is used for representing gradient change information of pixel points in a clear image corresponding to the noisy sample image, and the training set further includes reference gradient information of the clear image corresponding to the noisy sample image;

adjusting a network parameter of the gradient information prediction network based on a difference between the second prediction gradient information and the reference gradient information.

6. The method of claim 5, wherein the training set further comprises a sharp image corresponding to the noisy sample image;

the method further comprises the following steps:

and determining the reference gradient information by carrying out structure tensor analysis on a clear image corresponding to the noisy sample image.

7. The method of claim 5, wherein the training set further comprises a sharp image corresponding to the noisy sample image;

the method further comprises the following steps:

classifying pixel points in the noisy sample image according to the second prediction gradient information to obtain a second classification result corresponding to each pixel point in the noisy sample image;

based on the second classification result, performing classification convolution denoising on different types of pixel points through the classification denoising network to obtain a denoised image corresponding to the noisy sample image;

and adjusting the network parameters of the classified denoising network based on the difference between the denoised image corresponding to the noisy sample image and the clear image corresponding to the noisy sample image.

8. The method of any one of claims 1-7, wherein the gradient information comprises at least one of: gradient direction, gradient strength and gradient consistency.

9. An image denoising apparatus, comprising:

the first gradient information prediction module is used for carrying out image processing on a noisy image through a gradient information prediction network to obtain first prediction gradient information corresponding to the noisy image, wherein the first prediction gradient information is used for representing gradient change information of pixel points in a clear image corresponding to the noisy image;

the first classification module is used for classifying the pixel points in the image with noise according to the first prediction gradient information to obtain a first classification result corresponding to each pixel point in the image with noise;

and the first denoising module is used for denoising the image with the noise according to the first classification result to obtain a denoised image corresponding to the image with the noise.

10. The apparatus of claim 9, wherein the first classification module comprises:

and the third determining submodule is used for determining the first classification result according to the gradient classification corresponding to each pixel point.

11. The apparatus according to claim 9 or 10, wherein the first denoising module is specifically configured to:

12. The apparatus of claim 11, wherein the classification denoising network comprises a plurality of convolutional layers;

the first denoising module is specifically configured to:

13. The apparatus of any one of claims 9 to 12, further comprising:

the second gradient information prediction module is used for performing image processing on a noisy sample image included in a training set through the gradient information prediction network to obtain second prediction gradient information corresponding to the noisy sample image, wherein the second prediction gradient information is used for representing gradient change information of pixel points in a clear image corresponding to the noisy sample image, and the training set further includes reference gradient information of the clear image corresponding to the noisy sample image;

a first parameter adjustment module to adjust a network parameter of the gradient information prediction network based on a difference between the second predicted gradient information and the reference gradient information.

14. The apparatus according to claim 13, wherein the training set further comprises a sharp image corresponding to the noisy sample image;

the device further comprises:

and the structure tensor analysis module is used for carrying out structure tensor analysis on the clear image corresponding to the noisy sample image to determine the reference gradient information.

15. The apparatus according to claim 13, wherein the training set further comprises a sharp image corresponding to the noisy sample image;

the device further comprises:

the second denoising module is used for carrying out classification convolution denoising on different types of pixel points through the classification denoising network based on the second classification result to obtain a denoised image corresponding to the sample image with noise;

and the second network parameter adjusting module is used for adjusting the network parameters of the classified denoising network based on the difference between the denoised image corresponding to the noisy sample image and the clear image corresponding to the noisy sample image.

16. The apparatus of any one of claims 9-15, wherein the gradient information comprises at least one of: gradient direction, gradient strength and gradient consistency.

17. An electronic device, comprising:

a processor;

a memory for storing processor-executable instructions;

wherein the processor is configured to invoke the memory-stored instructions to perform the method of any one of claims 1 to 8.

18. A computer readable storage medium having computer program instructions stored thereon, which when executed by a processor implement the method of any one of claims 1 to 8.