Motion-based shape error concealment for object-based video

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 is assumed that the shape of the corrupted object at hand is in the form of a binary alpha plane and some of the shape data is missing due to channel errors. To conceal the corrupted shape, the decoder starts by assuming that the alpha plane changes in consecutive time instants can be described by a global motion model. This way, based on locally estimated global motion parameters, the decoder tries to conceal the corrupted alpha plane by global motion compensating the shape data from the previous time instant. Then, since not all alpha plane changes can be perfectly described by global motion, an additional local motion refinement is applied to deal with areas of the object that have significant motion.