Image deblurring method with prior information enhancement

Abstract

The application discloses a priori information enhanced image deblurring method, and belongs to the technical field of image processing. In the image edge information extraction and enhancement module, the clear edge information of an image is extracted by using a Canny operator, the blurred image features are combined with the edge information, information enhancement is carried out through Fourier transform and global pooling and other operations, an edge-enhanced feature map is obtained, and the edge high-frequency information of the image can be better recovered. In the foreground and background attention feature weighting module, the blurred image features are divided into foreground and background, the foreground and background features are obtained by multiplying the input blurred features, then the weighting map of the foreground and background features is obtained by using the pooling and convolution activation operation, the foreground and background information is weightedly fused, the structure information weighted feature map with different weights of the foreground and background is obtained, the position priori information of the foreground and the background is contained, and the deblurring can be better. The two modules are combined, so that the deblurring network has better generalization.

Description

Technical Field

[0001] This invention belongs to the field of image processing technology, and in particular relates to an image deblurring method that enhances prior information. Background Technology

[0002] Image deblurring is a crucial task in image processing and computer vision, aiming to recover a sharp image from a blurred one. In real life, blurred images not only affect people's visual perception but also negatively impact various fields such as medical imaging, surveillance, and autonomous driving. Therefore, research on image deblurring is particularly important. Blur can be mainly classified into several types: motion blur, focal blur, and Gaussian blur. Motion blur typically occurs when photographing fast-moving objects, resulting in motion blur in the image; focal blur occurs when the image is not properly focused, causing loss of detail; and Gaussian blur is a common image processing effect, often used for image smoothing. Image deblurring techniques have undergone several stages of development. Early deblurring methods were mostly based on classical mathematical models and filtering techniques, such as Wiener filtering and RL filtering. These methods performed well in handling simple blur and single-kernel blur, but often yielded poor results for complex blur situations. With the rapid development of computer vision and machine learning technologies, deep learning methods are increasingly being applied to image deblurring. Deep learning-based deblurring algorithms can handle complex blur patterns. Through training on large amounts of data, they learn the mapping relationship between blurred and sharp images, achieving end-to-end image deblurring. These methods demonstrate powerful capabilities in handling high dynamic range images, complex scenes, and various blur types. Deblurring technology has achieved significant results in multiple fields. In medical imaging, sharp images are crucial for lesion diagnosis. Image deblurring technology helps doctors observe subtle structures and lesions more accurately. In the field of security surveillance, blurry surveillance videos may lead to the loss of important evidence; deblurring technology can effectively improve the usability of surveillance images, helping law enforcement officers identify suspects.

[0003] Traditional image deblurring techniques are primarily based on prior knowledge and mathematical models. For example, Wiener filtering recovers an image by assuming statistical properties, while blind image deblurring based on deconvolution attempts to estimate both the blur kernel and the sharp image simultaneously. In real-world scenarios, a single image often contains different blur kernels. Estimating a single blur kernel is not conducive to image recovery. Furthermore, the solution space for solving the inverse problem of the sharp image is not uniquely determined, making it difficult to use blur kernels to solve for the sharp image. While these methods have a theoretical basis, they often perform poorly in practical applications, especially when dealing with complex blur types.

[0004] In recent years, the rapid development of deep learning has brought new opportunities to the field of image deblurring. Deblurring methods based on convolutional neural networks can be trained on a large number of blurred and sharp image pairs, adaptively learning image features to more effectively recover image details. These deep learning models can capture complex patterns and texture information that are difficult for traditional methods to handle, significantly improving deblurring results. However, despite the excellent performance of deep learning methods on many benchmarks, some challenges remain. Blurring in real-world scenes is often non-uniform, with different regions containing different blur kernels, resulting in poor recovery of image edge textures. Furthermore, model training requires a large amount of labeled data, and its generalization ability under different blur types and scenes still needs improvement. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention proposes an image deblurring scheme that enhances prior information.

[0006] The first aspect of this invention discloses an image deblurring method enhanced with prior information, the method comprising:

[0007] Step S1: Call the image edge information extraction and enhancement module, use the Canny operator to extract the edge information of the input blurred image, combine the blurred image features with the edge information, and enhance the edge information through Fourier transform and global pooling operations to obtain an edge enhancement feature map, which is used to restore the high-frequency edge information of the blurred image.

[0008] Step S2: Call the foreground and background attention feature weighting module to divide the blurred image features into foreground and background, multiply them by the blurred image features respectively to obtain foreground features and background features, obtain the weighted map of foreground features and background features through pooling and convolution activation operations, and then obtain the weighted feature map of foreground and background structural information with different weights through weighted fusion.

[0009] Step S3: Based on the edge enhancement feature map and the structural information weighted feature map of foreground and background with different weights, the input blurred image is deblurred using a deblurring network. The parameters of the deblurring network are obtained by pre-training with blurred-clear image pairs, and the deblurring network is composed of residual convolution.

[0010] According to the method of the first aspect of the present invention, a deblurring network is trained using a blurred-sharp image pair to obtain the optimal parameters of the deblurring network; wherein the blurred image in the blurred-sharp image pair is used as the input of the deblurring network, and a loss function is calculated using the sharp image and the deblurred image in the blurred-sharp image pair.

[0011] According to the method of the first aspect of the present invention, the input blurred image is preprocessed to obtain a blurred image feature map of size 3×H×W, where H×W is the image size and 3 is the number of image channels.

[0012] According to the method of the first aspect of the present invention, the Canny operator is used to extract edge information from the input blurred image feature map to obtain an edge information map. The edge information is combined with the blurred image features and an edge enhancement feature map is obtained through Fourier transform, convolutional activation, global average pooling, and inverse Fourier transform.

[0013] According to the method of the first aspect of the present invention, the edge information map is F_edge, and the edge information map F_edge and the blurred image feature map F_b are fused and input into the global pooling and Fourier transform branch to obtain the edge enhancement feature map F_s, and:

[0014] F_s=Conv(Avg(concate(F_b,F_edge)))+iFFT(Conv(FFT(concate(F_b,F_edge))))

[0015] Where Conv represents convolutional activation, Avg represents global average pooling, concate represents concatenation, iFFT represents inverse fast Fourier transform, FFT represents fast Fourier transform, and Conv(·) = Conv 1×1 (BatchNorm(ReLU(Conv 1×1 (·)))), ReLU represents the activation function, and BatchNorm represents batch normalization.

[0016] According to the method of the first aspect of the present invention, the blurred image features are divided into foreground features M1 and background features M2. Overfitting is eliminated through global pooling and max pooling. Global pooling extracts overall information and averages the information within each channel. After convolutional activation, a weighted map of the foreground and background features is obtained using the Sigmoid function. Then, weighted fusion is performed to obtain a weighted feature map F_w of the foreground and background with different weights.

[0017] F1=F_b*Sigmoid(Conv(concate(Avg(F_b*M1),Max(F_b*M1))))

[0018] F2=F_b*Sigmoid(Conv(concate(Avg(F_b*M2),Max(F_b*M2))))

[0019] F_w=F1+F2

[0020] The Sigmoid function maps feature values ​​to (0,1), Max represents global max pooling, F1 represents the foreground feature weighted map, and F2 represents the background feature weighted map.

[0021] According to the method of the first aspect of the present invention, an edge enhancement feature map F_s, a foreground and background structural information weighted feature map F_w with different weights, and a blurred image are input into a deblurring network to obtain a deblurred image.

[0022] A second aspect of this invention discloses an image deblurring system with prior information enhancement, characterized in that the system comprises an image edge information extraction and enhancement module, a foreground and background attention feature weighting module, and a deblurring network, wherein:

[0023] The image edge information extraction and enhancement module is configured to: extract the edge information of the input blurred image using the Canny operator, combine the blurred image features with the edge information, enhance the edge information through Fourier transform and global pooling operations, and obtain an edge enhancement feature map to recover the high-frequency edge information of the blurred image;

[0024] The foreground and background attention feature weighting module is configured to: divide the blurred image features into foreground and background, multiply them by the blurred image features respectively to obtain foreground features and background features, obtain a weighted map of foreground features and background features through pooling and convolution activation operations, and then obtain a weighted feature map of foreground and background structural information with different weights through weighted fusion.

[0025] The deblurring network is used to deblur the input blurred image based on the edge enhancement feature map and the structural information of the foreground and background with different weights. The parameters of the deblurring network are obtained by pre-training with blurred-sharp image pairs, and the deblurring network is composed of residual convolution.

[0026] A third aspect of this invention discloses an electronic device. The electronic device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the prior information enhancement image deblurring method described in the first aspect of this disclosure.

[0027] A fourth aspect of this invention discloses a computer-readable storage medium. The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the prior information enhancement image deblurring method described in the first aspect of this disclosure.

[0028] In summary, the technical solution proposed in this invention performs excellently in capturing edge frequency information and enhancing image structure understanding. It enhances the model's attention to different features, thereby solving the problems of insufficient edge information processing and image structure information perception in the prior art. It can better handle non-uniform blur in images, achieve better deblurring effect, and enhance the generalization of the model. Attached Figure Description

[0029] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0030] Figure 1 A flowchart illustrating an image deblurring method that enhances prior information;

[0031] Figure 2 This is a schematic diagram of the image edge information extraction and information enhancement module;

[0032] Figure 3 A schematic diagram of the foreground and background attention feature weighting module;

[0033] Figure 4 This is a schematic diagram of the structure of a deblurring network. Detailed Implementation

[0034] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0035] Image blurring is one of the most common image information loss phenomena in daily life. As an important information carrier, blurring not only affects human visual perception but also the accuracy of downstream computer vision tasks. In real-world scenarios, image blurring is non-uniform, with a very large solution space. This makes image deblurring an ill-posed problem with uncertain solutions, thus requiring further improvement in image restoration performance. Furthermore, the large solution space also hinders the generalization ability of image deblurring algorithms.

[0036] To address the limitations of existing methods, a structure-prior-based image deblurring scheme is proposed. This scheme excels in capturing edge frequency information and enhancing image structure understanding, thereby improving the model's attention to different features. It solves the problems of insufficient edge information processing and image structure information perception in existing methods, and can better handle non-uniform blur in images, achieving better deblurring results and enhancing the model's generalization ability.

[0037] The first aspect of this invention discloses an image deblurring method enhanced with prior information, the method comprising:

[0038] Step S1: Call the image edge information extraction and enhancement module, use the Canny operator to extract the edge information of the input blurred image, combine the blurred image features with the edge information, and enhance the edge information through Fourier transform and global pooling operations to obtain an edge enhancement feature map, which is used to restore the high-frequency edge information of the blurred image.

[0039] Step S2: Call the foreground and background attention feature weighting module to divide the blurred image features into foreground and background, multiply them by the blurred image features respectively to obtain foreground features and background features, obtain the weighted map of foreground features and background features through pooling and convolution activation operations, and then obtain the weighted feature map of foreground and background structural information with different weights through weighted fusion.

[0040] Step S3: Based on the edge enhancement feature map and the structural information weighted feature map of foreground and background with different weights, the input blurred image is deblurred using a deblurring network. The parameters of the deblurring network are obtained by pre-training with blurred-clear image pairs, and the deblurring network is composed of residual convolution.

[0041] According to the method of the first aspect of the present invention, a deblurring network is trained using a blurred-sharp image pair to obtain the optimal parameters of the deblurring network; wherein the blurred image in the blurred-sharp image pair is used as the input of the deblurring network, and a loss function is calculated using the sharp image and the deblurred image in the blurred-sharp image pair.

[0042] According to the method of the first aspect of the present invention, the input blurred image is preprocessed to obtain a blurred image feature map of size 3×H×W, where H×W is the image size and 3 is the number of image channels.

[0043] According to the method of the first aspect of the present invention, the Canny operator is used to extract edge information from the input blurred image feature map to obtain an edge information map. The edge information is combined with the blurred image features and an edge enhancement feature map is obtained through Fourier transform, convolutional activation, global average pooling, and inverse Fourier transform.

[0044] According to the method of the first aspect of the present invention, the edge information map is F_edge, and the edge information map F_edge and the blurred image feature map F_b are fused and input into the global pooling and Fourier transform branch to obtain the edge enhancement feature map F_s, and:

[0045] F_s=Conv(Avg(concate(F_b,F_edge)))+iFFT(Conv(FFT(concate(F_b,F_edge))))

[0046] Where Conv represents convolutional activation, Avg represents global average pooling, concate represents concatenation, iFFT represents inverse fast Fourier transform, FFT represents fast Fourier transform, and Conv(·) = Conv 1×1 (BatchNorm(ReLU(Conv 1×1 (·)))), ReLU represents the activation function, and BatchNorm represents batch normalization.

[0047] According to the method of the first aspect of the present invention, the blurred image features are divided into foreground features M1 and background features M2. Overfitting is eliminated through global pooling and max pooling. Global pooling extracts overall information and averages the information within each channel. After convolutional activation, a weighted map of the foreground and background features is obtained using the Sigmoid function. Then, weighted fusion is performed to obtain a weighted feature map F_w of the foreground and background with different weights.

[0048] F1=F_b*Sigmoid(Conv(concate(Avg(F_b*M1),Max(F_b*M1))))

[0049] F2=F_b*Sigmoid(Conv(concate(Avg(F_b*M2),Max(F_b*M2))))

[0050] F_w=F1+F2

[0051] The Sigmoid function maps feature values ​​to (0,1), Max represents global max pooling, F1 represents the foreground feature weighted map, and F2 represents the background feature weighted map.

[0052] According to the method of the first aspect of the present invention, an edge enhancement feature map F_s, a foreground and background structural information weighted feature map F_w with different weights, and a blurred image are input into a deblurring network to obtain a deblurred image.

[0053] A second aspect of this invention discloses an image deblurring system with prior information enhancement, characterized in that the system comprises an image edge information extraction and enhancement module, a foreground and background attention feature weighting module, and a deblurring network, wherein:

[0054] The image edge information extraction and enhancement module is configured to: extract the edge information of the input blurred image using the Canny operator, combine the blurred image features with the edge information, enhance the edge information through Fourier transform and global pooling operations, and obtain an edge enhancement feature map to recover the high-frequency edge information of the blurred image;

[0055] The foreground and background attention feature weighting module is configured to: divide the blurred image features into foreground and background, multiply them by the blurred image features respectively to obtain foreground features and background features, obtain a weighted map of foreground features and background features through pooling and convolution activation operations, and then obtain a weighted feature map of foreground and background structural information with different weights through weighted fusion.

[0056] The deblurring network is used to deblur the input blurred image based on the edge enhancement feature map and the structural information of the foreground and background with different weights. The parameters of the deblurring network are obtained by pre-training with blurred-sharp image pairs, and the deblurring network is composed of residual convolution.

[0057] First Embodiment

[0058] This embodiment relates to an image deblurring method enhanced with prior information, used to improve image deblurring performance. Compared to general image deblurring algorithms, the method in this embodiment mainly operates on the image edge information extraction and enhancement module and the foreground and background attention feature weighting module. For example... Figure 1 As shown, in the image edge information extraction and enhancement module, the Canny operator is used to extract clear edge information from the image. The blurred image features are combined with the edge information, and information enhancement is performed through operations such as Fourier transform and global pooling to obtain an edge enhancement feature map, which can better recover the high-frequency edge information of the image. In the foreground and background attention feature weighting module, the blurred image features are divided into foreground and background, and multiplied with the input blurred features to obtain foreground and background features. Then, pooling and convolution activation operations are used to obtain a weighted map of foreground and background features. The foreground and background information are then weighted and fused to obtain a structural information weighted feature map with different weights for the foreground and background, which includes prior positional information of the foreground and background, enabling better deblurring. Furthermore, combining these two modules gives the deblurring network better generalization ability.

[0059] Second Embodiment

[0060] This embodiment relates to the module composition of an image deblurring method that enhances prior information.

[0061] 1. Input module

[0062] The training data uses clear-blurred image pairs from an existing blurred image dataset, where the blurred images are used as input to the network and the clear images are used to calculate the loss function for deblurred images.

[0063] 2. Image edge information extraction and information enhancement module

[0064] The blurred image is processed using the Canny operator to extract edge information, resulting in an edge information map. This edge information map and the blurred feature map are then input into an image edge information extraction and enhancement module to obtain an enhanced edge feature map. The module structure is designed as follows: Figure 2 As shown.

[0065] 3. Foreground and Background Attention Feature Weighting Module

[0066] The input blurred image is divided into foreground features and background features, and then input into the foreground and background attention feature weighting module to obtain a foreground and background weighted feature map. The module's structure is as follows: Figure 3 As shown.

[0067] 4. Defuzzing network

[0068] The information features from the two modules mentioned above are fused and input into the deblurring network. Based on the training parameters of the blurred image pairs, the deblurred image is output. The deblurring network mainly consists of residual convolutions, such as... Figure 4 As shown.

[0069] Third Embodiment

[0070] This embodiment relates to the flowchart design of an image deblurring method enhanced by prior information. The aim is to enhance the model's attention to different features, thereby addressing the model's insufficient processing of edge information and perception of image structure information. This improves both the network's deblurring effect and the model's generalization ability. Therefore, the input of this device is a blurred image, and the output is a deblurred image. The algorithm flow is as follows:

[0071] Step S1: Obtain blurry-sharp image data pairs. After preprocessing the blurry image, obtain a blurry image feature map with an input size of 3×H×W, where H×W is the size of the input image and 3 is the number of channels of the input blurry image.

[0072] Step S2: Perform edge detection on the blurred image using the Canny operator to obtain a clearer edge map F_edge. Fuse the edge map F_edge and the blurred image feature map F_b, and input this into the global pooling and Fourier transform branches to fuse the edge features into the features of the blurred image, obtaining an edge enhancement feature map F_s, as shown below. Figure 2As shown, the overall calculation process is as follows:

[0073] F_s=Conv(Avg(concate(F_b,F_edge)))+iFFT(Conv(FFT(concate(F_b,F_edge))))

[0074] Where iFFT represents Inverse Fast Fourier Transform, FFT represents Fast Fourier Transform, and Conv represents the convolutional activation layer, which can be expressed as Conv(·) = Conv 1×1 (BatchNorm(ReLU(Conv 1×1 (·)))), concate means concatenating features along the channel dimension.

[0075] Step S3: Divide the blurred image features into foreground features M1 and background features M2. Multiply the foreground and background features by the blurred image feature map respectively to obtain the foreground and background parts of the features. Both parts are processed by global pooling and max pooling. Global max pooling highlights the most significant information in the feature map, thus mitigating overfitting to some extent. Global average pooling extracts the overall information of the feature map, averaging the information within each channel, which helps reduce the model's sensitivity to specific locations, thereby improving the model's generalization ability. After the convolutional activation layer, the Sigmoid function is used to obtain weighted features of the foreground and background. Then, the foreground and background features are weighted and fused to obtain a foreground-background weighted feature map F_w, as shown below. Figure 3 As shown, the overall calculation process is as follows:

[0076] F1=F_b*Sigmoid(Conv(concate(Avg(F_b*M1),Max(F_b*M1))))

[0077] F2=F_b*Sigmoid(Conv(concate(Avg(F_b*M2),Max(F_b*M2))))

[0078] F_w=F1+F2

[0079] The Sigmoid function maps eigenvalues ​​to the range (0,1), Avg represents global average pooling, and Max represents global max pooling.

[0080] Step S4: Input the edge enhancement feature F_s and foreground / background weighted feature F_w, along with the blurred image, into a database such as... Figure 4 The deblurring network shown yields a clear image after deblurring.

[0081] In summary, this invention proposes an image deblurring method enhanced with prior information. The proposed network, based on the deblurring network, adds feature enhancement and fusion of prior edge information and foreground / background position information. This enhances the image deblurring effect while improving the model's generalization ability. This invention proposes an image edge information extraction and enhancement module. Edge detection algorithms are used to extract edges from the blurred image. Fourier transform and convolutional pooling operations are used to fuse the frequency information of the edges, thereby improving the recovery of the edge contour information of the blurred image. This invention proposes a foreground and background attention feature weighting module. Blurred image features are divided into foreground and background features. The foreground and background are processed separately, and an attention mechanism is used to assign different weights, thereby effectively fusing the structural prior information of the foreground and background to improve the image deblurring effect.

[0082] A third aspect of this invention discloses an electronic device. The electronic device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the prior information enhancement image deblurring method described in the first aspect of this disclosure.

[0083] A fourth aspect of this invention discloses a computer-readable storage medium. The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the prior information enhancement image deblurring method described in the first aspect of this disclosure.

[0084] Please note that the technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments have been described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification. The above embodiments only illustrate several implementation methods of this application, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the invention patent. It should be pointed out that for those skilled in the art, several modifications and improvements can be made without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A prior information enhanced image deblurring method, characterized by, The method comprises: Step S1, calling an image edge information extraction and enhancement module, using a Canny operator to extract edge information of an input blurred image, combining blurred image features and edge information, performing edge information enhancement through Fourier transform and global pooling operation, and obtaining an edge-enhanced feature map for restoring edge high-frequency information of the blurred image; Step S2, calling a foreground and background attention feature weighting module, dividing the blurred image features into foreground and background, multiplying the foreground and background with the blurred image features respectively to obtain foreground features and background features, obtaining a weighted map of the foreground features and the background features through pooling and convolution activation operation, and then obtaining a foreground and background weighted feature map of structural information at different weights through weighted fusion; Step S3, inputting the edge-enhanced feature map and the foreground and background weighted feature map of structural information at different weights, using a deblurring network to perform deblurring processing on the input blurred image, and pre-training parameters of the deblurring network by using a blurred-clear image pair, and the deblurring network is composed of residual convolution.

2. The image deblurring method with prior information enhancement according to claim 1, characterized in that, The blurred-clear image pair is used to train the deblurring network to obtain optimal parameters of the deblurring network; wherein the blurred image in the blurred-clear image pair is used as input of the deblurring network, and the clear image and the deblurred image in the blurred-clear image pair are used to calculate a loss function.

3. The image deblurring method with prior information enhancement according to claim 2, characterized in that, The input blurred image is preprocessed to obtain a blurred image feature map with a size of 3×H×W, wherein H×W is the image size and 3 is the number of image channels.

4. The image deblurring method with prior information enhancement according to claim 3, characterized in that, The Canny operator is used to extract edge information from the input blurred image feature map to obtain an edge information map, the edge information is combined with the blurred image features, and an edge-enhanced feature map is obtained through Fourier transform, convolution activation, global average pooling, and inverse Fourier transform.

5. The prior information enhanced image deblurring method according to claim 4, characterized in that, The edge information map is F_edge, the edge information map F_edge and the blurred image feature map F_b are fused, input into a global pooling and Fourier transform branch, an edge-enhanced feature map F_s is obtained, and: F_s=Conv(Avg(concate(F_b,F_edge)))+iFFT(Conv(FFT(concate(F_b,F_edge)))) wherein Conv represents a convolution activation, Avg denotes a global average pooling, concate denotes a concatenation fusion, iFFT denotes an inverse fast Fourier transform, FFT denotes a fast Fourier transform, and Conv(·) = Conv 1×1 (BatchNorm(ReLU(Conv 1×1 (·)))), ReLU denotes an activation function, and BatchNorm denotes a batch normalization.

6. The image deblurring method with prior information enhancement according to claim 5, characterized in that, The blurred image features are divided into foreground and background, and are multiplied with the blurred image features respectively to obtain foreground features M1 and background features M2, global pooling and maximum pooling are used to eliminate overfitting, wherein global pooling extracts overall information and averages information within a channel; a Sigmoid function is used after convolution activation to obtain a weighted map of the foreground features and the background features, and then a foreground and background weighted feature map F_w of structural information at different weights is obtained through fusion: F1=F_b*Sigmoid(Conv(concate(Avg(F_b*M1),Max(F_b*M1)))) F2=F_b*Sigmoid(Conv(concate(Avg(F_b*M2),Max(F_b*M2)))) F_w=F1+F2 Wherein, the Sigmoid function maps the feature value to (0, 1), Max represents the global maximum pooling, F1 represents the foreground feature weighted graph, and F2 represents the background feature weighted graph.

7. The prior information enhanced image deblurring method of claim 6, wherein, The edge enhanced feature graph F_s, the foreground and background structure information weighted feature graph F_w at different weights, and the blurred image are input into the deblurring network to obtain a deblurred image.

8. A prior information enhanced image deblurring system, characterized by, The system comprises an image edge information extraction and enhancement module, a foreground and background attention feature weighting module, and a deblurring network. The image edge information extraction and enhancement module is configured to extract edge information of an input blurred image using a Canny operator, combine the blurred image features and the edge information, enhance the edge information through Fourier transform and global pooling operation, and obtain an edge enhanced feature graph for restoring the edge high frequency information of the blurred image. The foreground and background attention feature weighting module is configured to divide the blurred image features into foreground and background, multiply them with the blurred image features respectively to obtain foreground features and background features, obtain weighted graphs of the foreground features and the background features through pooling and convolution activation operation, and obtain the foreground and background structure information weighted feature graphs at different weights through weighted fusion. The deblurring network is used to input the edge enhanced feature graph and the foreground and background structure information weighted feature graph at different weights to perform deblurring processing on the input blurred image, the parameters of the deblurring network are obtained by pre-training of a blurred-clear image pair, and the deblurring network is composed of residual convolution.

9. An electronic device, comprising: The electronic device comprises a memory and a processor, the memory stores a computer program, and the processor implements the prior information enhanced image deblurring method of any one of claims 1-7 when executing the computer program.

10. A computer readable storage medium characterized by, The computer readable storage medium stores a computer program, and the computer program is executed by the processor to implement the prior information enhanced image deblurring method of any one of claims 1-7.

Images