Parameters optimization for a scalable multiple description coding scheme based on spatial subsampling

In this work, we design a multiple description coding (MDC) system for video streams moving from the concept of spatial MDC and introducing an efficient algorithm to obtain sub-streams that exploits some form of scalability. We first generate four subsequences by sub-sampling, then, from these four subsequences, by jointly coding two of them, we generate two descriptions. Finally, each description is compressed independently with the recent H.264/SVC video coding standard. In order to achieve some sort of scalability, for each description, we predict one of the original subsequences from the other one via inter layer prediction, thus generating a base layer, containing one subsequence, and an enhancement layer, containing the other one. In this paper, we present some results, varying the fraction of the total rate assigned to the base layer, in order to find the better value that guarantees the optimal performances in case not all the (sub)streams are received.