Multiple description motion compensation video coding for MPEG-4 FGS over lossy packet networks

A novel error resilience coding technique, named multiple description scalar quantization for fine granularity scalability (MDSQ-FGS), is presented to improve the temporal prediction efficiency of video coding over lossy packet networks and wireless channels. Despite easily adapting to channel bandwidth fluctuations, the coding efficiency (CE) of scalar coding techniques is low since only the base-layer is used in its motion prediction. But to achieve higher CE by using high quality reference frames in the enhancement-layer causes "drift" error propagation, caused by the mismatch between the reference frames used in encoding and decoding, when reference data are lost. The MDSQ-FGS video coding scheme is proposed to control drifting error without too much reduction in CE. Simulation results indicate that the proposed coder outperforms the normal MPEG-4 FGS coder for CE. It is also applicable for error prone wireless networks, for a mobile station moving between two access points.