A fast block matching algorthim for stereo correspondence

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

37 Scopus citations

Abstract

Stereo correspondence is one of the most active research areas in computer vision. An improved block matching approach for fast stereo correspondence is proposed, where the matching criterion is the sum of absolute difference (SAD). The approach tries to progressively confine the matching operation during the searching process, terminates unnecessary computations of matching criteria between the reference block in the left image and the ineligible candidate block in the right image, and eliminates the ineligible block as early as possible while ensuring the optimal disparity of each pixel. The performance of the new algorithm is evaluated by carrying out a theoretical analysis, and by comparing it with the stereo correspondence method based on the standard block matching. Simulation results demonstrate that the disparities obtained by this algorithm are identical to that using standard block matching method, and a reduction of over 55% in computational cost is achieved.

Original languageEnglish
Title of host publication2008 IEEE International Conference on Cybernetics and Intelligent Systems, CIS 2008
PublisherIEEE Computer Society
Pages38-41
Number of pages4
ISBN (Print)9781424416745
DOIs
StatePublished - 2008
Event2008 IEEE International Conference on Cybernetics and Intelligent Systems, CIS 2008 - Chengdu, China
Duration: 21 Sep 200824 Sep 2008

Publication series

Name2008 IEEE International Conference on Cybernetics and Intelligent Systems, CIS 2008

Conference

Conference2008 IEEE International Conference on Cybernetics and Intelligent Systems, CIS 2008
Country/TerritoryChina
CityChengdu
Period21/09/0824/09/08

Keywords

  • Block matching, disparity
  • Computer vision
  • SAD
  • Stereo correspondence

Fingerprint

Dive into the research topics of 'A fast block matching algorthim for stereo correspondence'. Together they form a unique fingerprint.

Cite this