A second-order approach for blind motion deblurring by normalized l1 regularization

  • Zedong Chen
  • , Faming Fang
  • , Yingying Xu
  • , Chaomin Shen*
  • *Corresponding author for this work

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

1 Scopus citations

Abstract

We propose a second-order approach for blind motion deblurring. Our idea is to define an energy functional, and the convolution kernel corresponds to the minimum of the functional. After the kernel is obtained, the problem is solved by existing non-blind deconvolution algorithms. To avoid that the minimizer of energy functional does not correspond to the unblurred image, which is often encountered in many algorithms, in the literature the normalized l1 norm regularization term for the original or first-order gradient image was adopted. We further extend the idea using the second-order gradient image, which is the main novelty of the paper. This method favours a piecewise linear transition in the unblurred image, and thus efficiently attenuates the staircase and ring effects in the original or first-order case. Comparing with other stateof- the-art algorithms, the proposed method is effective in estimating the blur-kernel and restoring the unblurred image.

Original languageEnglish
Title of host publicationAdvances in Multimedia Information Processing – 17th Pacific-Rim Conference on Multimedia, PCM 2016, Proceedings
EditorsEnqing Chen, Yun Tie, Yihong Gong
PublisherSpringer Verlag
Pages296-305
Number of pages10
ISBN (Print)9783319488950
DOIs
StatePublished - 2016
Event17th Pacific-Rim Conference on Multimedia, PCM 2016 - Xi’an, China
Duration: 15 Sep 201616 Sep 2016

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume9917 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference17th Pacific-Rim Conference on Multimedia, PCM 2016
Country/TerritoryChina
CityXi’an
Period15/09/1616/09/16

Keywords

  • Blind motion deblurring
  • Normalized l norm
  • Secondorder gradient
  • Staircase effects

Fingerprint

Dive into the research topics of 'A second-order approach for blind motion deblurring by normalized l1 regularization'. Together they form a unique fingerprint.

Cite this