Blind image restoration and segmentation via decoupled adaptive Mumford Shah model

Zhangqin Jiang, Fengwen Mi*, Zeyang Dou

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

A new model that can simultaneously do blind restoration and segmentation task is proposed in the paper. The new model belongs to the variant of Mumford Shah model. In order to promote the computational efficiency, the restoration part and segmentation part are decoupled from the original model. The blind image restoration part is based on the variable exponent regularizer to accurately estimate both piecewise constant point spread functions and smooth point spread functions. The segmentation part is the explicit edge indicator function obtained from the original model. The new model can be efficiently solved using split bregman framework. Numerical experiments show that the new algorithm produces promising results and robust to noise.

Original languageEnglish
Title of host publicationSelected Papers of the Chinese Society for Optical Engineering Conferences held October and November 2016
EditorsHesheng Chen, Jianyu Wang, Jialing Le, Jianda Shao, Yueguang Lv
PublisherSPIE
ISBN (Electronic)9781510610118
DOIs
Publication statusPublished - 2017
EventChinese Society for Optical Engineering Conferences, CSOE 2016 - Jinhua, Suzhou, Chengdu, Xi'an, and Wuxi, China
Duration: 1 Nov 2016 → …

Publication series

NameProceedings of SPIE - The International Society for Optical Engineering
Volume10255
ISSN (Print)0277-786X
ISSN (Electronic)1996-756X

Conference

ConferenceChinese Society for Optical Engineering Conferences, CSOE 2016
Country/TerritoryChina
CityJinhua, Suzhou, Chengdu, Xi'an, and Wuxi
Period1/11/16 → …

Keywords

  • Blind deconvolution
  • Mumford Shah functional
  • Point spread function
  • Segmentation
  • Split bregman iteration

Fingerprint

Dive into the research topics of 'Blind image restoration and segmentation via decoupled adaptive Mumford Shah model'. Together they form a unique fingerprint.

Cite this