Sequential error concealment via canonical correlation analysis

Fan Wen, Liang Junli, Ye Xin, Min Li

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

1 Scopus citations

Abstract

In this paper, we propose a new error concealment algorithm for video/image based on canonical correlation analysis (CCA). Motivated by the Intra prediction in H.264/AVC, it is reasonable to assume that there is a strong spatial correlation relationship between the lost regions and their known adjacent regions. Based on the above idea, we use CCA to estimate a correlation projection matrix which utilizes the loss of macro block adjacent spatial information, then we use the projection matrix and the adjacent region to estimate missing pixel area. In addition, in order to use the spatial information efficiently, we apply the neighbor-embedding-type weight into the aforementioned CCA model. Experimental results demonstrate that the proposed method improves the subjective and objective image quality to a large extent in comparison with other existing methods.

Original languageEnglish
Title of host publication2015 IEEE International Conference on Signal Processing, Communications and Computing, ICSPCC 2015
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781479989188
DOIs
StatePublished - 25 Nov 2015
Event5th IEEE International Conference on Signal Processing, Communications and Computing, ICSPCC 2015 - Ningbo, Zhejiang, China
Duration: 19 Sep 201522 Sep 2015

Publication series

Name2015 IEEE International Conference on Signal Processing, Communications and Computing, ICSPCC 2015

Conference

Conference5th IEEE International Conference on Signal Processing, Communications and Computing, ICSPCC 2015
Country/TerritoryChina
CityNingbo, Zhejiang
Period19/09/1522/09/15

Keywords

  • canonical correlation analysis
  • Error concealment
  • image processing

Fingerprint

Dive into the research topics of 'Sequential error concealment via canonical correlation analysis'. Together they form a unique fingerprint.

Cite this