Error Resilient Multiple Description Video Coding over Wireless Ad-hoc Networks.

摘要

Providing reliable video communications over wireless ad-hoc networks is becoming increasingly important as these networks become widely deployed in military, homeland defense, and disaster recovery applications. However, wireless ad-hoc networks impose great challenges to support such applications due to the highly dynamic network topology and the unreliable wireless channels. Given the error-prone nature of the wireless ad-hoc networks and the vulnerability of compressed video to packet losses, it is critical to enhance error resilience of video transmission over such lossy networks.;In this dissertation, we investigate approaches to enhance error robustness of delivered video based on multiple description coding (MDC) and a multipath transport (MPT) framework. We first investigate error concealment techniques at the decoder to improve the reconstructed video with transmission errors. The main idea is to utilize the redundancy among multiple video descriptions to provide better concealment for intra and inter frames on a macroblock (MB) basis. Since the concealment methods may cause error propagation among descriptions, we further introduce a rate-distortion optimized (RDO) mode selection method at the encoder to enhance resilience, which is motivated by the recursive optimal per-pixel estimate (ROPE) approach. This method estimates the end-to-end distortion for MDC, considering the network conditions and multiple state recovery, and uses the estimated distortion to select the optimal coding mode and thus reduces error propagation due to packet losses. The RDO mode selection method for MD C requires knowledge of network conditions, which is time-varying and not easy to obtain in wireless ad-hoc networks.;We observe that routing messages available in the standard routing protocols indicate route changes and link failures of the network and can be used to track potential packet losses. Thus, we propose an approach that estimates the packet losses based on the routing messages and selects reference frames at the encoder accordingly. We establish a model to estimate the packet loss probability of each packet based on the routing messages and network parameters, without an additional feedback channel or extra overhead. By identifying the corrupted transmitted frames and avoiding using them as reference in multiple description coding, we reduce error propagation and improve the delivered video quality.