Local‐global stereo matching algorithm

Abstract

Purpose

In order to solve the problem that dynamic programming (DP) stereo algorithm is slow, a local‐global stereo matching algorithm is presented. In this method, local and global stereo algorithms are united to apply to the stereo vision field.

Design/methodology/approach

Local and DP stereo algorithms are used. Multi‐resolution images are obtained by pyramid algorithm and matching costs are gained by using eight‐stripe correlation window.

Findings

The time of disparity space image creation is greatly reduced because matching costs are computed by eight‐stripe window. Candidate ground control points (GCPs) are gained in the low‐resolution image. The total time of gaining GCPs is decreased in the extreme. Whole algorithm has a high matching‐rate and a fast speed.

Research limitations/implications

The time of the whole algorithm is still a little bigger than local algorithm.

Practical implications

The findings can be used in the stereo vision system equipped with planet vehicle.

Originality/value

Local and global stereo algorithms are united to apply to the stereo vision field. Matching costs are gained by using eight‐stripe correlation window.

Citation

Zhao, J., Yu, S. and Cai, H. (2006), "Local‐global stereo matching algorithm", Aircraft Engineering and Aerospace Technology, Vol. 78 No. 4, pp. 289-292. https://doi.org/10.1108/17488840610675564