A new segment-based stereo matching using graph cuts

Abstract

In the paper, we formulate a new energy function followed by the use of graph cuts to refine the disparity map which takes segment as node. Firstly, the robust disparity plane fitting is modeled and the method of Singular Value Decomposition (SVD) is used to solve least square. In order to ensure reliable pixel sets for the segment, we filter out outliers through three main rules, namely; cross-checking, judging reliable area and measuring the distance between previous disparity to the computed disparity plane. Secondly, we apply improve hierarchical clustering algorithm to merge neighbor. Finally, the final disparity map is obtained. Experimental results demonstrate that our approach is effective in improving the state of the art. © 2010 IEEE.