| This dissertation is concerned with enhancing the error resilience of video transmission over lossy networks, such as the Internet and wireless networks. The major objective is to balance the tradeoff between efficiency and robustness within a rate-distortion (RD) framework, so that the best reconstruction quality is achieved under the prescribed bit rate budget and network constraints.; The overall quality of video transmission over lossy networks is an intricate function of several factors including the source and channel coding schemes, the network status, and the decoder's recovery actions. This fact motivates the derivation of the proposed “end-to-end” approach.; The first part of the dissertation focuses on the system performance analysis and the derivation of efficient algorithms to estimate the end-to-end distortion. The live video coding and pre-compressed video streaming scenarios are addressed separately, due to the fundamental difference in the level of network awareness during compression, which hence poses different challenges and requires different solutions. A recursive optimal per-pixel estimate (ROPE) is proposed for calculating, at the encoder, the expected end-to-end distortion of live video coding and transmission over lossy networks. For the streaming of fixed pre-compressed video, a first order distortion estimation (FODE) algorithm is derived, which is applicable to virtually all source encoding techniques, network settings, varying channel conditions, and specific error-resilience schemes. Both algorithms take into account the effects of quantization, packet loss, spatial and temporal error propagation, and error concealment. They feature high accuracy and low complexity, and are compatible with video coding standards.; Given the above capacity for accurate end-to-end distortion analysis, the second part of the dissertation is concerned with balancing the tradeoff between efficiency and robustness within a rate-distortion framework. The control parameters in a given system are optimally selected to explicitly and directly minimize the end-to-end distortion. We first propose a technique of RD-optimized adaptive intra-update at the MB level to optimize the robustness of the live video transmission. The technique is then extended to handle the standard layered coding system, including enhancement of both encoder-based and decoder-based error-resilience measures. We finally propose an RD-optimized adaptation solution for the robust streaming of pre-compressed video. Substantial gains in the simulation results demonstrate the superiority of the proposed end-to-end RD-optimized approaches. |
