A Two-Branch Convolution Residual Network for Image Compressive Sensing
Deep learning has made great progress in image compressive sensing (CS) tasks recently, and several CS models based on it have achieved superior performance. In practice, sensing the entire image requires huge memory and computational effort. Although the block-based CS method can effectively realiz...
Main Authors: | , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
IEEE
2020-01-01
|
Series: | IEEE Access |
Subjects: | |
Online Access: | https://ieeexplore.ieee.org/document/8938811/ |
id |
doaj-7ec0b528019d45dd99eb6285f7942d50 |
---|---|
record_format |
Article |
spelling |
doaj-7ec0b528019d45dd99eb6285f7942d502021-03-30T01:12:37ZengIEEEIEEE Access2169-35362020-01-0181705171410.1109/ACCESS.2019.29613698938811A Two-Branch Convolution Residual Network for Image Compressive SensingChenquan Gan0https://orcid.org/0000-0002-0453-5630Xiaoqin Yan1https://orcid.org/0000-0001-6517-0121Yunfeng Wu2https://orcid.org/0000-0002-2740-2224Zufan Zhang3https://orcid.org/0000-0001-5315-2065School of Communication and Information Engineering, Chongqing University of Post and Telecommunications, Chongqing, ChinaSchool of Communication and Information Engineering, Chongqing University of Post and Telecommunications, Chongqing, ChinaSchool of Communication and Information Engineering, Chongqing University of Post and Telecommunications, Chongqing, ChinaSchool of Communication and Information Engineering, Chongqing University of Post and Telecommunications, Chongqing, ChinaDeep learning has made great progress in image compressive sensing (CS) tasks recently, and several CS models based on it have achieved superior performance. In practice, sensing the entire image requires huge memory and computational effort. Although the block-based CS method can effectively realize image sensing, it will cause block effects that severely decrease the reconstruction performance. To this end, this paper proposes a two-branch convolution residual network for image compressive sensing (denoted as TCR-CS), which mainly consists of a two-branch convolution autoencoder network and a residual network. Specifically, the two-branch convolution autoencoder network senses the entire image through multiple scale convolutional filters to obtain measurements. For better CS reconstruction, the image is preliminarily reconstructed by the deconvolution decoder network, and then the residual network is used to optimize the pre-reconstructed image. Through the end-to-end training, all networks can be jointly optimized. Finally, experimental results demonstrate that the proposed TCR-CS method is superior to existing state-of-the-art CS methods in terms of structural similarity, reconstruction performance and visual quality at different measurement rates.https://ieeexplore.ieee.org/document/8938811/Image compressive sensingtwo-branch convolutionresidual networkstructural similarityreconstruction performancevisual quality |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Chenquan Gan Xiaoqin Yan Yunfeng Wu Zufan Zhang |
spellingShingle |
Chenquan Gan Xiaoqin Yan Yunfeng Wu Zufan Zhang A Two-Branch Convolution Residual Network for Image Compressive Sensing IEEE Access Image compressive sensing two-branch convolution residual network structural similarity reconstruction performance visual quality |
author_facet |
Chenquan Gan Xiaoqin Yan Yunfeng Wu Zufan Zhang |
author_sort |
Chenquan Gan |
title |
A Two-Branch Convolution Residual Network for Image Compressive Sensing |
title_short |
A Two-Branch Convolution Residual Network for Image Compressive Sensing |
title_full |
A Two-Branch Convolution Residual Network for Image Compressive Sensing |
title_fullStr |
A Two-Branch Convolution Residual Network for Image Compressive Sensing |
title_full_unstemmed |
A Two-Branch Convolution Residual Network for Image Compressive Sensing |
title_sort |
two-branch convolution residual network for image compressive sensing |
publisher |
IEEE |
series |
IEEE Access |
issn |
2169-3536 |
publishDate |
2020-01-01 |
description |
Deep learning has made great progress in image compressive sensing (CS) tasks recently, and several CS models based on it have achieved superior performance. In practice, sensing the entire image requires huge memory and computational effort. Although the block-based CS method can effectively realize image sensing, it will cause block effects that severely decrease the reconstruction performance. To this end, this paper proposes a two-branch convolution residual network for image compressive sensing (denoted as TCR-CS), which mainly consists of a two-branch convolution autoencoder network and a residual network. Specifically, the two-branch convolution autoencoder network senses the entire image through multiple scale convolutional filters to obtain measurements. For better CS reconstruction, the image is preliminarily reconstructed by the deconvolution decoder network, and then the residual network is used to optimize the pre-reconstructed image. Through the end-to-end training, all networks can be jointly optimized. Finally, experimental results demonstrate that the proposed TCR-CS method is superior to existing state-of-the-art CS methods in terms of structural similarity, reconstruction performance and visual quality at different measurement rates. |
topic |
Image compressive sensing two-branch convolution residual network structural similarity reconstruction performance visual quality |
url |
https://ieeexplore.ieee.org/document/8938811/ |
work_keys_str_mv |
AT chenquangan atwobranchconvolutionresidualnetworkforimagecompressivesensing AT xiaoqinyan atwobranchconvolutionresidualnetworkforimagecompressivesensing AT yunfengwu atwobranchconvolutionresidualnetworkforimagecompressivesensing AT zufanzhang atwobranchconvolutionresidualnetworkforimagecompressivesensing AT chenquangan twobranchconvolutionresidualnetworkforimagecompressivesensing AT xiaoqinyan twobranchconvolutionresidualnetworkforimagecompressivesensing AT yunfengwu twobranchconvolutionresidualnetworkforimagecompressivesensing AT zufanzhang twobranchconvolutionresidualnetworkforimagecompressivesensing |
_version_ |
1724187499186618368 |