Video quality enhancement through End-to-End distortion optimization and enriched video traces

Transmission of video streams over a network with varying delay and loss probability characteristics poses a great challenge in ensuring the reception of visually acceptable video quality at the receivers. As the availability of the important video data may be affected by the latency and the loss of video packets, considering these factors, the challenge is then to provide adequate decoded video quality. In this thesis, the root causes of video quality degradation are investigated.;An optimization model is proposed that mathematically describes the distortions of each traffic stream in terms of the availability of the data partition packets along with the network loss probability. The developed distortion model will further combine the distortions from all the traffic streams in the system and minimize the distortion for all the streams. The optimization model is implemented in the queue of the Edge Router to the Internet and the optimization program is activated once the delay and loss probability exceed the predetermined thresholds. The outcome of the optimization model is to provide a priority order for all the streams where the anticipated E2E distortions for all the streams are minimized. The packets in the queues will be then prioritized and reordered according to the deployment order determined by the optimization program.;Results from the optimizer simulation runs demonstrate improvements in the received video quality, measured by a quantitative metric: the Peak Signal-to-Noise Ratio (PSNR) value, in all of the traffic flows in comparison to the PSNR values for the traffic flows in the simulation runs where no optimization was applied. The improvements for different traffic flows are varied as the optimizer awards preference to the traffic flows whose anticipated received PSNR values are measured to be at risk during the time of reordering and deployment. The results obtained from the simulation runs fully support the E2E distortion optimization mathematical model with respect to enhancing the quality for all the streams when the optimization was applied.;In summary, the contributions of this thesis are (i) a new video trace format containing granular level distortion parameters, (ii) a mathematical formulation for E2E distortion in video streams and (iii) a novel distortion optimization method which enhances the E2E video quality through prioritization of streams deployment are the main contribution of this thesis.;Furthermore, a new video trace format is introduced to reflect detailed level of video information. In addition to the parameters reflected on the conventional video traces, the proposed trace format includes granular information at each Data Partition level for every video Slice. This new trace format also includes the contributing distortion parameters (The "K" values) which represent the contributing distortion of each Data Partition. The "K" factors are utilized by the algorithms in the Edge Routers to intelligently calculate the End-to-End (E2E) predicted distortion values for each stream.