Fast blind deconvolution using a deeper sparse patch-wise maximum gradient prior
作者：
Highlights：
• We propose a novel patch-wise maximum gradient (PMG) prior for blind deblurring.
• We verify the validity of the PMG prior on both the dataset and theory.
• We propose a new L0-regularized PMG prior, and develop an optimization scheme.
• Our method achieves excellent results in calculating cost and deblurring quality.
• Our method can perform deblurring on both natural and domain-specific images.
摘要
•We propose a novel patch-wise maximum gradient (PMG) prior for blind deblurring.•We verify the validity of the PMG prior on both the dataset and theory.•We propose a new L0-regularized PMG prior, and develop an optimization scheme.•Our method achieves excellent results in calculating cost and deblurring quality.•Our method can perform deblurring on both natural and domain-specific images.
论文关键词：Blind deblurring,Patch-wise maximum gradient prior,L0-regularized prior
论文评审过程：Received 6 February 2020, Revised 2 August 2020, Accepted 16 October 2020, Available online 31 October 2020, Version of Record 4 November 2020.
论文官网地址：https://doi.org/10.1016/j.image.2020.116050
