Motion compensated shape error concealment

摘要: The introduction of Video Objects (VOs) is one the innovations MPEG-4. /spl alpha/-plane a VO defines its shape at given instance in time and hence determines boundary texture. In packet-based networks, shape, motion, texture are subject to loss. While there has been considerable attention paid concealment motion errors, little done field error concealment. this paper we propose post-processing technique that uses compensated information previously received alpha/-plane. proposed approach based on matching segments current frame previous frame. This achieved by finding maximally smooth vector field. After matched boundary, missing pieces reconstructed compensation. Experimental results demonstrating performance method, comparing it with weighted side method presented.