Error-resilient rate shaping for video streaming over packet-loss networks

Video streaming over packet-loss networks faces the challenges that the networks are error-prone, transmission bandwidth is limited and fluctuating, the user device capabilities vary, and networks are heterogeneous. These challenges necessitate the need for smart adaptation of the precoded video. The focus of the thesis is error-resilient rate shaping for streaming precoded video over packet-loss networks. Given the packet-loss characteristic of the networks, the precoded video consists of channel-coded as well as source-coded bits. Error-resilient rate shaping is a filtering process that adapts the bit rates of the precoded video, in order to deliver the best video quality given the network condition at the time of delivery. We first illustrate “baseline rate shaping (BRS)” of the proposed error-resilient rate shaping as a baseline. Having introduced BRS with coarse decisions in rate adaptation, more sophisticated error-resilient rate shaping is proposed for layer-coded videos, namely, the enhancement layer video and the base layer video. “Fine-grained rate shaping (FGRS)” is proposed for streaming the enhancement layer video, and “error-concealment aware rate shaping (SCARS)” is proposed for streaming the base layer video. FGRS and SCARS are formulated as rate-distortion (R-D) optimization problems. A two-stage R-D optimization approach is proposed to solve the R-D optimization problem in a fast and accurate manner. FGRS makes use of the fine granularity property of the MPEG-4 fine-granularity-scalability bitstream and outperforms ad-hoc unequal packet-loss protection methods. SCARS takes into account error concealment (EC) performed at the receiver to deliver the part of precoded video that cannot be EC-reconstructed well. Frame dependency due to predictive coding and/or temporal EC is also considered in SCARS by means of feedback from the receiver. Experiments are conducted under various channel conditions and for various types of the video to demonstrate the effectiveness of the proposed scheme. Finally, we see that network conditions are needed in optimizing the streaming performance. In the last part of the thesis, we focus on modeling the video traffic so that we may use the syntactic traffic to probe the network to determine the network condition and optimize the proposed error-resilient rate shaping accordingly.