Tuning stereo image matching with stereo video sequence processing

摘要: Algorithms for stereo video image processing typicaly assume that the various tasks; calibration, static matching, and egomotion are independent black boxes. In particular, task of computing disparity estimates is normally performed independently ongoing environmental recovery processes. Can information from these processes be exploited in notoriously hard problem field estimation? Here we explore use feedback model being constructed to stereopsis task. A prior estimate used seed stereomatching process within a probabilistic framework. Experimental results on simulated real data demonstrate potential approach.