Super-resolution reconstruction method and system for image

Abstract

The invention discloses a super-resolution reconstruction method and system for an image. The method is based on a channel attention mechanism and a position attention mechanism of an image. Featureson each image block are extracted at different depths and then fused, and the importance of the characteristic channel information and the position information is effectively identified; and since thefeatures of different depths make different contributions to the image super-resolution, the method and the system make full use of and fuses the features of different depths, the obtained image features retain a larger receptive field and better detail features at the same time, and the super-resolution effect of the obtained image is better after reconstruction based on the features. Besides, the relation between the features is fully mined so that the performance bottleneck problem caused by the too deep network can be alleviated, the feature representation capability of the image is great, and the robustness can be realized.

Description

technical field

[0001] The invention belongs to the field of computer vision, and more specifically relates to a method and system for image super-resolution reconstruction. Background technique

[0002] Although the early difference-based image super-resolution methods are simple and efficient, their effects in practical applications are greatly limited. Recently, methods based on deep convolutional neural networks have achieved performance beyond traditional image super-resolution methods. The earliest method based on deep convolutional neural network proposes a three-layer image super-resolution network, which includes four parts: shallow feature extraction, nonlinear mapping, reconstruction and upsampling. For the first time, deep learning Introduced into the image super-resolution task. Based on the above network structure, the image super-resolution network is gradually deepened. By introducing a residual learning structure, the super-resolution network can be deepen...

Examples

Embodiment 1

[0056] An image super-resolution reconstruction method, such as figure 1 shown, including the following steps:

[0057] S1. Dividing the low-resolution image to be processed into blocks to obtain multiple image blocks;

[0058] Specifically, in this embodiment, the low-resolution image to be processed is sequentially and repeatedly intercepted image blocks with a size of H×W in a sliding window manner. Specifically, the low-resolution RGB image to be processed has repeated image blocks with a size of H×W cut from left to right and from top to bottom in the form of a sliding window.

[0059] S2. Based on the channel attention mechanism and position attention mechanism of the image, the features on each image block are extracted at different depths, and the feature maps of different depths corresponding to each image block are obtained;

[0060] Preferably, the following steps are included:

[0061] S21. Using a convolutional neural network to extract shallow features of the ...

Embodiment 2

[0089] An image super-resolution system, such as image 3 As shown, including: image interception module 1, feature extraction module 2, image reconstruction module 3 and image recombination module 4;

[0090] Wherein, the image interception module 1 is used for dividing the low-resolution image to be processed into blocks to obtain a plurality of image blocks, and output them to the feature extraction module 2;

[0091] The feature extraction module 2 is used for image-based channel attention mechanism and position attention mechanism, extracts the features of each image block input by the image interception module at different depths, obtains feature maps of different depths corresponding to each image block, and outputs to In the image reconstruction module 3; preferably, the feature extraction module includes a shallow network unit 21 and a deep network unit 22; wherein the shallow network unit 21 is used to extract the shallow features of the image block using a convoluti...

Embodiment 3

[0098] A storage medium, when a computer reads the instructions, it causes the computer to execute the image super-resolution reconstruction method provided in Embodiment 1 of the present invention.

technical field

[0001] The invention belongs to the field of computer vision, and more specifically relates to a method and system for image super-resolution reconstruction. Background technique

[0002] Although the early difference-based image super-resolution methods are simple and efficient, their effects in practical applications are greatly limited. Recently, methods based on deep convolutional neural networks have achieved performance beyond traditional image super-resolution methods. The earliest method based on deep convolutional neural network proposes a three-layer image super-resolution network, which includes four parts: shallow feature extraction, nonlinear mapping, reconstruction and upsampling. For the first time, deep learning Introduced into the image super-resolution task. Based on the above network structure, the image super-resolution network is gradually deepened. By introducing a residual learning structure, the super-resolution network can be deepen...

Examples

Embodiment 1

[0056] An image super-resolution reconstruction method, such as figure 1 shown, including the following steps:

[0057] S1. Dividing the low-resolution image to be processed into blocks to obtain multiple image blocks;

[0058] Specifically, in this embodiment, the low-resolution image to be processed is sequentially and repeatedly intercepted image blocks with a size of H×W in a sliding window manner. Specifically, the low-resolution RGB image to be processed has repeated image blocks with a size of H×W cut from left to right and from top to bottom in the form of a sliding window.

[0059] S2. Based on the channel attention mechanism and position attention mechanism of the image, the features on each image block are extracted at different depths, and the feature maps of different depths corresponding to each image block are obtained;

[0060] Preferably, the following steps are included:

[0061] S21. Using a convolutional neural network to extract shallow features of the ...

Embodiment 2

[0089] An image super-resolution system, such as image 3 As shown, including: image interception module 1, feature extraction module 2, image reconstruction module 3 and image recombination module 4;

[0090] Wherein, the image interception module 1 is used for dividing the low-resolution image to be processed into blocks to obtain a plurality of image blocks, and output them to the feature extraction module 2;

[0091] The feature extraction module 2 is used for image-based channel attention mechanism and position attention mechanism, extracts the features of each image block input by the image interception module at different depths, obtains feature maps of different depths corresponding to each image block, and outputs to In the image reconstruction module 3; preferably, the feature extraction module includes a shallow network unit 21 and a deep network unit 22; wherein the shallow network unit 21 is used to extract the shallow features of the image block using a convoluti...

Embodiment 3

[0098] A storage medium, when a computer reads the instructions, it causes the computer to execute the image super-resolution reconstruction method provided in Embodiment 1 of the present invention.

Images