Abstract
With the advances in video compression and networking techniques, the last ten years have witnessed an explosive growth of video applications over the Internet. However, the service model of the current best-effort network was never engineered to handle video traffic and, as a result, video applications still suffer from varying and unpredictable network conditions, in terms of bandwidth, packet loss and delay. To address these problems, a lot of innovative techniques have been proposed and researched. Among them, scalable video coding is a promising one to cope with the dynamics of the available bandwidth and heterogeneous terminals. This work aims at improving the efficacy of scalable video transport over IP networks. In this work, we first propose an optimal interleaving scheme combined with motion-compensated fine granularity scalability video source coding and unequal loss protection schemes, under an imposed delay constraint. The network is modeled as a packet-loss channel with random delays. The motion compensation prediction, ULP allocation and the depth of the interleaver are jointly optimized based on the network status and the delay constraint. We then proceed to investigate the multiple path transport technique. A unified approach which incorporates adaptive motion compensation prediction, multiple description coding and unequal multiple path allocation, is proposed to improve both the robustness and error resilience property of the video coding and transmission system, while the delivered video quality is improved simultaneously. To analytically investigate the efficacy of error resilient transport schemes for progressively encoded sources, including unequal loss protection, best-effort and FEC transport schemes, we develop evaluation and optimization approaches for these transport schemes. In this part of the work, the network is modeled as an M/D/1/K queue, and then a comprehensive queueing analysis is provided. Armed with these results, the efficacy of these transport schemes for progressively encoded sources are investigated and compared.