ROBUST VERGENCE WITH CONCURRENT IDENTIFICATION OF OCCLUSION AND SPECULAR HIGHLIGHTS

The early detection, classification, and elimination of occlusion and specular highlights which occur frequently in real scenes are of utmos t importance. However, most of vergence algorithms which have been pro posed in recent years do not take into consideration the effect of occ lusion and specular highlights present in real scenes. In this paper, we propose an active exploratory vergence method based on concurrent c rosscorrelation of multiscale stereo images and an adaptive feedback s trategy using weighted normalized cross-correlation to detect and dist inguish occlusion and specular highlights concurrently during the verg ence process. It not only improves computational efficiency but also l eads to the exploration of a better viewing direction and position tha t overcome the problem created by occlusion and specular highlights. T he robustness of the proposed method against occlusion and specular hi ghlights is demonstrated with real scenes. (C) 1995 Academic Press, In c.