Fairness-oriented multi-stream rate adaptation using scalable video coding

In the delivery of video services like video on-demand, IP-TV, sport broadcasting, as well as real-time streaming, the end-user expectation is to receive the best feasible quality independently of the particular video complexity, even in the presence of packet losses. In this scenario, rate adaptation is required to optimize the overall quality, whereas fairness is an important issue that has to be addressed. In this paper we propose a multi-stream rate adaptation framework with reference to the scalable video coding (SVC) extension of the H.264/AVC standard with medium grain scalability (MGS). We first define a general discrete multi-objective problem with the aim to maximize the sum of assigned rates, while minimizing the differences among the expected distortions, under a total bit-rate constraint. A single-objective problem formulation is then derived by applying a continuous relaxation. Finally, a simplified continuous semi-analytical model that accurately estimates the rate–distortion relationship for both error-free channel and packet-erasure channel is also proposed, which allows us to derive an optimal and low-complexity procedure to solve the relaxed problem. Unequal erasure protection (UXP) is also considered and designed to suitably shape the rate–distortion relationship for different values of RTP packet-loss rate. The numerical results show the goodness of our framework in terms of error gap between the relaxed and its related discrete solution, and the significant performance improvement achieved with respect to an equal-rate adaptation scheme.