Quality driven system design for video delivery over dynamically changing networks

This thesis focuses on optimizing video quality for two types of video applications over dynamically changing networks: video multicast in wireless networks (suitable for delivering popular events to a group of mobile users in the same wireless coverage area), and individual on-demand streaming over HTTP protocol.;Cooperative communication is a promising research direction to integrate with the idea of video multicast. Video transmission over wireless channels suffers from random fluctuation of the channel condition. This thesis introduces a novel and efficient approach for user cooperation in wireless video multicast using Randomized Distributed Space Time Codes (R-DSTC), in which the sender first transmits the source packets, and the sender and receivers that have received all source packets then generate and send the parity packets simultaneously using R-DSTC. As more parity packets are delivered, more receivers can recover all source packets and join the parity packet generation and transmission. The proposed system can support significantly higher video rates and correspondingly higher quality of decoded video, than prior approaches in the literature.;Dynamic adaptive streaming over HTTP (DASH) is currently the dominant approach in commercial on-demand video streaming applications. In a Dash system, the streaming server stores multiple representations (with different rates) of the same video segment and the streaming client requests the representation for the next segment based on its playback buffer status and the estimated bandwidth. To optimize perceptual video quality, it is important to understand the visual effect of the significant variation of the quantization step-size due to the change in the video rate. We conducted several subjective quality experiments for video sequences with non-periodic temporal variation patterns. We present subjective test results and examine how does the overall quality under non-periodic variation relate to the qualities of individual periodic components, and how does it relate to the instantaneous qualities at different time points. We propose two quality models based on our observations and data analysis. We introduce an algorithm for the client to request the video representation for each segment so that the overall perceptual video quality is maximized over a certain period based on the proposed quality model. Another visual quality experiment was conducted to evaluate videos resulting from our proposed adaptation algorithm against other commonly used algorithms. The videos resulting from the proposed algorithm are consistently ranked higher than the benchmark algorithms.