A Syntax-Based Approach to Multiple Description Video Coding

The paper describes an efficient method to achieve Multiple Description Coding (MDC) for video streaming applications. The idea is to use a JPEG-like syntax to represent an ordered sequence of the coefficients, alternately distributed among the two descriptors. The ordering procedure allows an efficient interpolation of the missing coefficients, and then an almost seamless reconstruction from a single descriptor. As compared to other approaches to MDC, the proposed strategy is characterized by an extremely low computational burden, while ensuring comparable overhead and good quality of single-descriptor video. We will consider the basic scheme, where only two descriptions are taken into account, bearing in mind that an extension to more descriptions is feasible with small effort, thanks to the extremely light nature of the encoder and decoder structure.