Motion-based shape error concealment for object-based video

摘要: In this paper, an original motion-based shape error concealment technique, especially useful for object-based video applications in error-prone environments such as mobile networks, is proposed. It assumed that the of corrupted object at hand form a binary alpha plane and some data missing due to channel errors. To conceal shape, decoder starts by assuming changes consecutive time instants can be described global motion model. This way, based on locally estimated parameters, tries compensating from previous instant. Then, since not all perfectly motion, additional local refinement applied deal with areas have significant motion.