Error resilient video coding using multiple description motion compensation

Abstract

A new approach for multiple description video coding using motion compensated prediction is proposed. The temporal predictor is designed to provide a desired trade-off between prediction gain and mismatch between the predictions obtained at the encoder and at the decoder when only one description is available. The mismatch signal is explicitly coded to avoid error propagation. The system can achieve a wide range of redundancy, by varying the predictor and the quantizer for the mismatch signal.