Syntax and discontinuity based error concealment for compressed video in a packet environment

This paper addresses the problem of error resilience of compressed video. The errors considered are lost packets which are generally caused by network congestion in a packet switched network. We propose distributing the bits from a single slice among several packets so that the loss of a packet will result in only one or a few errors in any one slice. The loss of the packet is expected to be detected by the depacketizer and the bits thus lost are treated as erasures. We proceed to generate candidates for corrupted slices and test their syntactical integrity and their spatial smoothness. The smoothest candidate for a slice which is syntatically correct is chosen as the corrected slice. The scheme need not add overhead to the bitstream, bits will insert delay. A parameter of our scheme is the number of bits from a single slice inserted into one packet (M). Results of delay simulations over the parameter M are given. Complexity is also simulated over the parameter M. The improvement in performance objectively and subjectively is shown.