Deblurring by Realistic Blurring

被引:325
作者
Zhang, Kaihao [1 ]
Luo, Wenhan [2 ]
Zhong, Yiran [1 ,4 ]
Ma, Lin [2 ]
Stenger, Bjorn [3 ]
Liu, Wei [2 ]
Li, Hongdong [1 ,4 ]
机构
[1] Australian Natl Univ, Canberra, ACT, Australia
[2] Tencent AI Lab, Bellevue, WA USA
[3] Rakuten Inst Technol, Bangalore, Karnataka, India
[4] ACRV, Brisbane, Qld, Australia
来源
2020 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR) | 2020年
关键词
D O I
10.1109/CVPR42600.2020.00281
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Existing deep learning methods for image deblurring typically train models using pairs of sharp images and their blurred counterparts. However, synthetically blurring images do not necessarily model the genuine blurring process in real-world scenarios with sufficient accuracy. To address this problem, we propose a new method which combines two GAN models, i.e., a learning-to-Blur GAN (BGAN) and learning-to-DeBlur GAN (DBGAN), in order to learn a better model for image deblurring by primarily learning how to blur images. The first model, BGAN, learns how to blur sharp images with unpaired sharp and blurry image sets, and then guides the second model, DBGAN, to learn how to correctly deblur such images. In order to reduce the discrepancy between real blur and synthesized blur, a relativistic blur loss is leveraged. As an additional contribution, this paper also introduces a Real-World Blurred Image (RWBI) dataset including diverse blurry images. Our experiments show that the proposed method achieves consistently superior quantitative performance as well as higher perceptual quality on both the newly proposed dataset and the public GOPRO dataset.
引用
收藏
页码:2734 / 2743
页数:10
相关论文
共 46 条
[1]   To Learn Image Super-Resolution, Use a GAN to Learn How to Do Image Degradation First [J].
Bulat, Adrian ;
Yang, Jing ;
Tzimiropoulos, Georgios .
COMPUTER VISION - ECCV 2018, PT VI, 2018, 11210 :187-202
[2]   A Neural Approach to Blind Motion Deblurring [J].
Chakrabarti, Ayan .
COMPUTER VISION - ECCV 2016, PT III, 2016, 9907 :221-235
[3]   Total variation blind deconvolution [J].
Chan, TF ;
Wong, CK .
IEEE TRANSACTIONS ON IMAGE PROCESSING, 1998, 7 (03) :370-375
[4]  
Chen HJ, 2018, IEEE INT CONF COMPUT
[5]   Fast Motion Deblurring [J].
Cho, Sunghyun ;
Lee, Seungyong .
ACM TRANSACTIONS ON GRAPHICS, 2009, 28 (05) :1-8
[6]   Dynamic Scene Deblurring with Parameter Selective Sharing and Nested Skip Connections [J].
Gao, Hongyun ;
Tao, Xin ;
Shen, Xiaoyong ;
Jia, Jiaya .
2019 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2019), 2019, :3843-3851
[7]  
Goldman A, 2012, IEEE C EVOL COMPUTAT
[8]  
Gupta A, 2010, LECT NOTES COMPUT SC, V6311, P171, DOI 10.1007/978-3-642-15549-9_13
[9]   Deep Residual Learning for Image Recognition [J].
He, Kaiming ;
Zhang, Xiangyu ;
Ren, Shaoqing ;
Sun, Jian .
2016 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2016, :770-778
[10]  
Hirsch M, 2011, IEEE I CONF COMP VIS, P463, DOI 10.1109/ICCV.2011.6126276