Hyperspectral restoration employing low rank and 3D total variation regularization

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

14 Scopus citations

Abstract

This paper presents a novel mixed-noise removal method by employing low rank constraint and 3-D total variation regularization for hyperspectral image (HSI) restoration. The main idea of the proposed method is based on the assumption that the spectra in HSI lie in the same low rank subspace and both spatial and spectral domains exhibit the property of piecewise smoothness. The low rank property of HSI is exploited by the nuclear norm, while the spectral-spatial smoothness is explored by 3D total variation (3DTV) which is defined as a combination of 2-D spatial TV and 1-D spectral TV of the HSI cube. Finally, the proposed restoration model is effectively solved by alternating direction method of multipliers (ADMM). Experimental results on simulated HSI dataset validate the superior performance of the proposed method.

Original languageEnglish
Title of host publicationPIC 2016 - Proceedings of the 2016 IEEE International Conference on Progress in Informatics and Computing
EditorsYinglin Wang, Yaoru Sun
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages326-329
Number of pages4
ISBN (Electronic)9781509034833
DOIs
StatePublished - 15 Jun 2017
Event4th IEEE International Conference on Progress in Informatics and Computing, PIC 2016 - Shanghai, China
Duration: 23 Dec 201625 Dec 2016

Publication series

NamePIC 2016 - Proceedings of the 2016 IEEE International Conference on Progress in Informatics and Computing

Conference

Conference4th IEEE International Conference on Progress in Informatics and Computing, PIC 2016
Country/TerritoryChina
CityShanghai
Period23/12/1625/12/16

Keywords

  • 3D total variation
  • ADMM
  • Hyperspectral image
  • Low rank property
  • Restoration

Fingerprint

Dive into the research topics of 'Hyperspectral restoration employing low rank and 3D total variation regularization'. Together they form a unique fingerprint.

Cite this