Skip to main navigation Skip to search Skip to main content

Fast disparity estimation and mode decision for multi-view video coding

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

Abstract

Disparity estimation and mode decisions are key techniques in multi-view video coding (MVC) which could improve the compression efficiency when the computational complexity increasing greatly. Based on Kalman filtering, a novel fast disparity estimation and mode decision algorithm is presented in this paper. We firstly built a autoregressive (AR) model of disparity vectors on the basis of spatio-temporal correlation so as to achieve a preliminary result of disparity estimation. Furthermore, the Kalman filter is utilized to optimize and improve the estimation speed. Moreover, an effective reliability judgment method for mode prediction is presented, with which, a more precious mode prediction result can be obtained and the selected range of coding mode is effectively reduced to achieve low complexity mode decision. The experimental results show that the computational complexity is significantly reduced while the compression efficiency is still maintained.

Original languageEnglish
Title of host publicationProceedings - 2014 International Conference on Medical Biometrics, ICMB 2014
PublisherIEEE Computer Society
Pages195-199
Number of pages5
ISBN (Print)9781479940141
DOIs
StatePublished - 2014
Externally publishedYes
Event2014 International Conference on Medical Biometrics, ICMB 2014 - Shenzhen, Guangdong, China
Duration: 30 May 20141 Jun 2014

Publication series

NameProceedings - 2014 International Conference on Medical Biometrics, ICMB 2014

Conference

Conference2014 International Conference on Medical Biometrics, ICMB 2014
Country/TerritoryChina
CityShenzhen, Guangdong
Period30/05/141/06/14

Keywords

  • Kalman filtering
  • disparity estimation
  • mode decision
  • multi-view video coding

Fingerprint

Dive into the research topics of 'Fast disparity estimation and mode decision for multi-view video coding'. Together they form a unique fingerprint.

Cite this