Non-uniform motion deblurring with Kernel grid regularization

作者：

Highlights：

•

• Using a Kernel mapping flow as a constraint to estimate Kernels for spatially varying blur model is proposed.

• Detect and optimize the blur Kernel relies on the Earth mover’s distance.

• Ink Dot Diffusion based model as a recurrent framework to expand the optimized region.

• An effective method that solves the specific problem with a wild background is proposed.

摘要

•Using a Kernel mapping flow as a constraint to estimate Kernels for spatially varying blur model is proposed.•Detect and optimize the blur Kernel relies on the Earth mover’s distance.•Ink Dot Diffusion based model as a recurrent framework to expand the optimized region.•An effective method that solves the specific problem with a wild background is proposed.

论文关键词：Non-uniform deblurring,Blind deconvolution,Total variation regularization,Kernel mapping

论文评审过程：Received 27 January 2017, Revised 3 December 2017, Accepted 4 December 2017, Available online 12 December 2017, Version of Record 19 December 2017.

论文官网地址：https://doi.org/10.1016/j.image.2017.12.002