Optimized Source-Channel Coding of Video Signals in Packet Loss Environments

Abstract

A novel predictive joint source-channel video coding scheme is proposed and its superiority against standard video coding is demonstrated in environments with heavy packet loss. The strength of the scheme stems from the fact that it explicitly takes into account the two modes of operation (packet loss or no packet loss) at the decoder and optimizes the corresponding reconstruction filters together with the the prediction filter at the encoder simultaneously. As a result, the prediction coefficient tends to be much smaller than both the correlation coefficient between two corresponding frames and what standard video coding techniques use (i.e., 1), thereby leaving most of the inter-frame correlation intact and increasing the error resilience.