Video streaming over lossy/variable networks by means of multiple description

In this paper we propose a system based on multiple description coding (MD) to stream video over lossy or variable bandwidth networks. MD coding can be integrated in several ways into existing state-of-the-art video codecs, as MPEG or the new H.264. The proposed scheme is based on polyphase downsampling in the pixel domain, followed by independent coding of descriptions by means of existing video codecs. Other schemes are cited for comparison. For the proposed scheme, several enhancements are discussed: frames expansion, joint encoding and joint decoding. We compare the coding efficiency and the robustness to losses and errors with respect to standard and error-resilient single description coding (SD).