Recurrent Conditional Generative Adversarial Network for Image Deblurring
Nowadays, there is an increasing demand for images with high definition and fine textures, but images captured in natural scenes usually suffer from complicated blurry artifacts, caused mostly by object motion or camera shaking. Since these annoying artifacts greatly decrease image visual quality, d...
Main Authors: | , , |
---|---|
Format: | Article |
Language: | English |
Published: |
IEEE
2019-01-01
|
Series: | IEEE Access |
Subjects: | |
Online Access: | https://ieeexplore.ieee.org/document/8585006/ |
id |
doaj-6982d1b2c1a446969138e33b115d1f5c |
---|---|
record_format |
Article |
spelling |
doaj-6982d1b2c1a446969138e33b115d1f5c2021-03-29T22:07:01ZengIEEEIEEE Access2169-35362019-01-0176186619310.1109/ACCESS.2018.28888858585006Recurrent Conditional Generative Adversarial Network for Image DeblurringJing Liu0https://orcid.org/0000-0003-4690-1886Wanning Sun1Mengjie Li2School of Electrical and Information Engineering, Tianjin University, Tianjin, ChinaSchool of Electrical and Information Engineering, Tianjin University, Tianjin, ChinaSchool of Electrical and Information Engineering, Tianjin University, Tianjin, ChinaNowadays, there is an increasing demand for images with high definition and fine textures, but images captured in natural scenes usually suffer from complicated blurry artifacts, caused mostly by object motion or camera shaking. Since these annoying artifacts greatly decrease image visual quality, deblurring algorithms have been proposed from various aspects. However, most energy-optimization-based algorithms rely heavily on blur kernel priors, and some learning-based methods either adopt pixel-wise loss function or ignore global structural information. Therefore, we propose an image deblurring algorithm based on a recurrent conditional generative adversarial network (RCGAN), in which the scale-recurrent generator extracts sequence spatio–temporal features and reconstructs sharp images in a coarse-to-fine scheme. To thoroughly evaluate the global and local generator performance, we further propose a receptive field recurrent discriminator. Besides, the discriminator takes blurry images as conditions, which helps to differentiate reconstructed images from real sharp ones. Last but not least, since the gradients are vanishing when training the generator with the output of the discriminator, a progressive loss function is proposed to enhance the gradients in back propagation and to take full advantage of discriminative features. Extensive experiments prove the superiority of RCGAN over state-of-the-art algorithms both qualitatively and quantitatively.https://ieeexplore.ieee.org/document/8585006/Image deblurringconditional generative adversarial networkreceptive field recurrentcoarse-to-fine |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Jing Liu Wanning Sun Mengjie Li |
spellingShingle |
Jing Liu Wanning Sun Mengjie Li Recurrent Conditional Generative Adversarial Network for Image Deblurring IEEE Access Image deblurring conditional generative adversarial network receptive field recurrent coarse-to-fine |
author_facet |
Jing Liu Wanning Sun Mengjie Li |
author_sort |
Jing Liu |
title |
Recurrent Conditional Generative Adversarial Network for Image Deblurring |
title_short |
Recurrent Conditional Generative Adversarial Network for Image Deblurring |
title_full |
Recurrent Conditional Generative Adversarial Network for Image Deblurring |
title_fullStr |
Recurrent Conditional Generative Adversarial Network for Image Deblurring |
title_full_unstemmed |
Recurrent Conditional Generative Adversarial Network for Image Deblurring |
title_sort |
recurrent conditional generative adversarial network for image deblurring |
publisher |
IEEE |
series |
IEEE Access |
issn |
2169-3536 |
publishDate |
2019-01-01 |
description |
Nowadays, there is an increasing demand for images with high definition and fine textures, but images captured in natural scenes usually suffer from complicated blurry artifacts, caused mostly by object motion or camera shaking. Since these annoying artifacts greatly decrease image visual quality, deblurring algorithms have been proposed from various aspects. However, most energy-optimization-based algorithms rely heavily on blur kernel priors, and some learning-based methods either adopt pixel-wise loss function or ignore global structural information. Therefore, we propose an image deblurring algorithm based on a recurrent conditional generative adversarial network (RCGAN), in which the scale-recurrent generator extracts sequence spatio–temporal features and reconstructs sharp images in a coarse-to-fine scheme. To thoroughly evaluate the global and local generator performance, we further propose a receptive field recurrent discriminator. Besides, the discriminator takes blurry images as conditions, which helps to differentiate reconstructed images from real sharp ones. Last but not least, since the gradients are vanishing when training the generator with the output of the discriminator, a progressive loss function is proposed to enhance the gradients in back propagation and to take full advantage of discriminative features. Extensive experiments prove the superiority of RCGAN over state-of-the-art algorithms both qualitatively and quantitatively. |
topic |
Image deblurring conditional generative adversarial network receptive field recurrent coarse-to-fine |
url |
https://ieeexplore.ieee.org/document/8585006/ |
work_keys_str_mv |
AT jingliu recurrentconditionalgenerativeadversarialnetworkforimagedeblurring AT wanningsun recurrentconditionalgenerativeadversarialnetworkforimagedeblurring AT mengjieli recurrentconditionalgenerativeadversarialnetworkforimagedeblurring |
_version_ |
1724192173657686016 |