A new approach for multiple description video coding using motioncompensated prediction is proposed. The temporal predictor is designedto provide a desired trade-off between prediction gain and mismatchbetween the predictions obtained at the encoder and at the decoder whenonly one description is available. The mismatch signal is explicitlycoded to avoid error propagation. The system can achieve a wide range ofredundancy, by varying the predictor and the quantizer for the mismatchsignal
展开▼