Learning Fully Convolutional Networks for Iterative Non-blind Deconvolution

被引：94

作者：

Zhang, Jiawei ^{[1
,3
]}

Pan, Jinshan ^{[2
]}

Lai, Wei-Sheng ^{[3
]}

Lau, Rynson W. H. ^{[1
]}

Yang, Ming-Hsuan ^{[3
]}

机构：

[1] City Univ Hong Kong, Dept Comp Sci, Hong Kong, Hong Kong, Peoples R China

[2] Dalian Univ Technol, Sch Math Sci, Dalian, Peoples R China

[3] Univ Calif, Elect Engn & Comp Sci, Merced, CA 95340 USA

来源：

30TH IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2017) | 2017年

基金：

中国国家自然科学基金;

关键词：

FIELDS;

D O I：

10.1109/CVPR.2017.737

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

In this paper, we propose a fully convolutional network for iterative non-blind deconvolution. We decompose the non-blind deconvolution problem into image denoising and image deconvolution. We train a FCNN to remove noise in the gradient domain and use the learned gradients to guide the image deconvolution step. In contrast to the existing deep neural network based methods, we iteratively deconvolve the blurred images in a multi-stage framework. The proposed method is able to learn an adaptive image prior, which keeps both local (details) and global (structures) information. Both quantitative and qualitative evaluations on the benchmark datasets demonstrate that the proposed method performs favorably against state-of-the-art algorithms in terms of quality and speed.

引用

页码：6969 / 6977

页数：9