A real-time spatio-temporal stereo matching for road applications

This paper presents a real-time approach for matching stereo images acquired by a stereo sensor embedded in a moving vehicle. The new method consists of matching edge points extracted from stereo images using the temporal relationship, which exists between consecutive stereo pairs. Matching a current stereo pair takes into account the matching results of the preceding stereo pair. The method looks first for what we call matching control edge points (MCEPs) based on spatio-temporal matching of edge curves of consecutive stereo pairs. Dynamic programming is considered for matching edge points of the stereo images. The MCEPs drive the optimal path of the dynamic programming. The proposed approach has been tested on virtual and real stereo image sequences and the results are satisfactory.