Model-Based Deep Network for Single Image Deraining
For current learning-based single image deraining methods, deraining networks are usually designed based on a simplified linear additive rain model, which may not only cause unreal synthetic rainy images for both training and testing datasets, but also adversely affect the applicability and generali...
Main Authors: | , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
IEEE
2020-01-01
|
Series: | IEEE Access |
Subjects: | |
Online Access: | https://ieeexplore.ieee.org/document/8955865/ |
id |
doaj-d1a4d95fbcd2478d97902130ddcda47f |
---|---|
record_format |
Article |
spelling |
doaj-d1a4d95fbcd2478d97902130ddcda47f2021-03-30T03:10:36ZengIEEEIEEE Access2169-35362020-01-018140361404710.1109/ACCESS.2020.29655458955865Model-Based Deep Network for Single Image DerainingPengyue Li0https://orcid.org/0000-0003-1156-2423Jiandong Tian1https://orcid.org/0000-0002-5027-1381Yandong Tang2https://orcid.org/0000-0003-3805-7654Guolin Wang3https://orcid.org/0000-0003-1974-4432Chengdong Wu4https://orcid.org/0000-0001-5574-6932Faculty of Robot Science and Engineering, Northeastern University, Shenyang, ChinaState Key Laboratory of Robotics, Shenyang Institute of Automation, Chinese Academy of Sciences, Shenyang, ChinaState Key Laboratory of Robotics, Shenyang Institute of Automation, Chinese Academy of Sciences, Shenyang, ChinaState Key Laboratory of Robotics, Shenyang Institute of Automation, Chinese Academy of Sciences, Shenyang, ChinaFaculty of Robot Science and Engineering, Northeastern University, Shenyang, ChinaFor current learning-based single image deraining methods, deraining networks are usually designed based on a simplified linear additive rain model, which may not only cause unreal synthetic rainy images for both training and testing datasets, but also adversely affect the applicability and generality of corresponding networks. In this paper, we use the screen blend model of Photoshop as the nonlinear rainy image decomposition model. Based on this model, we design a novel channel attention U-DenseNet for rain detection and a residual dense block for rain removal. The detection sub-network not only adjusts channel-wise feature responses by our novel channel attention block to pay more attention to learn the rain map, but also combines the context information with the precise localization by the U-DenseNet to promote pixel-wise estimation accuracy. After rain detection, we use the nonlinear model to get a coarse rain-free image, and then introduce a deraining refinement subnetwork consisted of the residual dense block to obtain a fine rain-free image. For training our network, we apply the nonlinear rain model to synthesize a benchmark dataset called as RITD. It contains 3200 triplets of rainy images, rain maps, and clean background images. Our extensive quantitative and qualitative experimental results show that our method outperforms several state-of-the-art methods on both synthetic and real images.https://ieeexplore.ieee.org/document/8955865/Rain removalnonlinear rain modelchannel attention U-DenseNetresidual dense blockimage restoration |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Pengyue Li Jiandong Tian Yandong Tang Guolin Wang Chengdong Wu |
spellingShingle |
Pengyue Li Jiandong Tian Yandong Tang Guolin Wang Chengdong Wu Model-Based Deep Network for Single Image Deraining IEEE Access Rain removal nonlinear rain model channel attention U-DenseNet residual dense block image restoration |
author_facet |
Pengyue Li Jiandong Tian Yandong Tang Guolin Wang Chengdong Wu |
author_sort |
Pengyue Li |
title |
Model-Based Deep Network for Single Image Deraining |
title_short |
Model-Based Deep Network for Single Image Deraining |
title_full |
Model-Based Deep Network for Single Image Deraining |
title_fullStr |
Model-Based Deep Network for Single Image Deraining |
title_full_unstemmed |
Model-Based Deep Network for Single Image Deraining |
title_sort |
model-based deep network for single image deraining |
publisher |
IEEE |
series |
IEEE Access |
issn |
2169-3536 |
publishDate |
2020-01-01 |
description |
For current learning-based single image deraining methods, deraining networks are usually designed based on a simplified linear additive rain model, which may not only cause unreal synthetic rainy images for both training and testing datasets, but also adversely affect the applicability and generality of corresponding networks. In this paper, we use the screen blend model of Photoshop as the nonlinear rainy image decomposition model. Based on this model, we design a novel channel attention U-DenseNet for rain detection and a residual dense block for rain removal. The detection sub-network not only adjusts channel-wise feature responses by our novel channel attention block to pay more attention to learn the rain map, but also combines the context information with the precise localization by the U-DenseNet to promote pixel-wise estimation accuracy. After rain detection, we use the nonlinear model to get a coarse rain-free image, and then introduce a deraining refinement subnetwork consisted of the residual dense block to obtain a fine rain-free image. For training our network, we apply the nonlinear rain model to synthesize a benchmark dataset called as RITD. It contains 3200 triplets of rainy images, rain maps, and clean background images. Our extensive quantitative and qualitative experimental results show that our method outperforms several state-of-the-art methods on both synthetic and real images. |
topic |
Rain removal nonlinear rain model channel attention U-DenseNet residual dense block image restoration |
url |
https://ieeexplore.ieee.org/document/8955865/ |
work_keys_str_mv |
AT pengyueli modelbaseddeepnetworkforsingleimagederaining AT jiandongtian modelbaseddeepnetworkforsingleimagederaining AT yandongtang modelbaseddeepnetworkforsingleimagederaining AT guolinwang modelbaseddeepnetworkforsingleimagederaining AT chengdongwu modelbaseddeepnetworkforsingleimagederaining |
_version_ |
1724183909827084288 |