Error resilient video streaming over lossy networks

The objective of this thesis is to provide improved error-resiliency for video streaming over error-prone networks. In order to achieve the best video quality of service in the presence of channel disturbances, we first investigate post-processing error concealment methods that operate at the client side. A new multiframe recovery principle is proposed to provide improved error concealment capabilities. This approach analyzes the propagation of lost macroblocks (MBs) into succeeding frames, and these lost MBs are recovered in such a way that the error propagation is minimized.; In addition to post-processing approaches, an error-resilient source coding technique, called Multiple Description Coding (MDC), is addressed in this thesis. We develop a post-processing algorithm by using redundant information of a video signal to improve the quality for multiple description coded video. In this proposed research, we modify the decoder operation to estimate the lost description by utilizing the spatial/temporal redundancy in video signals. Besides, in order to solve the prediction mismatch problem, we propose two novel motion-compensated predictive coding methods for multiple description video coding. Finally, we address the challenges of optimally streaming multiple description coded video over multiple independent paths as well as a single path. Two different solutions are provided. If multiple paths are available, the proposed MDC-aware packet discarding method can effectively prevent the network from dropping both descriptions simultaneously. However, if only a single path can be used, we propose a novel packet scheduling method for MDC to minimize the end-to-end distortion.