Optimized mesh routing with intermediate recovery for error resilient delivery of MD coded image/video content

摘要

Multiple Description (MD) source coding is a technique that breaks a media stream into equally important sub streams which can be sent over different paths for protection against wireless channel errors. In this paper, we explore the possibility of sending these descriptions through different paths that merge at some specific intermediate nodes, where the corrupted descriptions are recovered from the uncorrupted ones, thereby increasing the quality of received image/video at the destination. To quantify the gain with intermediate recovery, we first devise an analytic model with simplifying assumptions on the network system for quantifying end-to-end distortion of MD coded data as a function of path parameters, and demonstrate that on a Lena image transmitted over long multipath routes, one intermediate recovery stage offers up to 9.2% reduction in distortion compared to the traditional multipath transport. Next, accounting the random network topology we formulate mesh route construction as a cross-layer optimization problem to balance between end-to-end packet delivery delay and distortion. Since this problem is highly complex, we propose two alternative delay/distortion minimization heuristics. Further, a jointly delay and distortion optimizing genetic algorithm based meta-heuristic route construction technique is suggested for networks with highly varying link quality. NS2-based simulations of a realistic network scenario demonstrate that, in terms of peak signal-to-noise ratio the intermediate recovery approach results in substantial improvement, close to 15 dB, in quality of video delivery.