Graph-cut versus belief-propagation stereo on real-world images

Abstract

This paper deals with a comparison between the performance of graph cuts and belief propagation stereo matching algorithms over long real-world and synthetics sequences. The results following different preprocessing steps as well as the running times are investigated. The usage of long stereo sequences allows us to better understand the behavior of the algorithms and the preprocessing methods, as well as to have a more realistic evaluation of the algorithms in the context of a vision-based Driver Assistance System (DAS). © 2009 Springer-Verlag Berlin Heidelberg.