Priority and Path Diversity Based Video Streaming over Best Effort Networks

Abstract

Successful delivery of important parts of video sequences is expected to maintain continuous playback at an acceptable video quality. In this thesis, priority and path diversity-based schemes are proposed in order to enhance the viewer’s experience by reducing the delay and loss rates of important video packets that might occur due to routing to a congested path. The introduced schemes classify video frames into three priority levels based on their importance. In addition, the proposed schemes maintain a priority queuing system at each router to reduce the delay experienced by video frames with high priority and hence to reduce their probability of dropping. The performance of the proposed schemes is evaluated through simulations in comparison to each other as well as to a best effort-based model. A priority-based queuing model is introduced in which each router maintains a priority queuing system without the use of path diversity. Moreover, we introduced a priority-based path diversity (PBPD) scheme that integrates the priority-based queuing model with path diversity. The performance of the three models is studied in terms of network related metrics such as frame loss rates and end-to-end delay. Moreover, the performance of the three models is compared from the end users’ perspective in terms of the peak-signal-to-noise-ratio (PSNR). In addition, a buffer occupancy metric is used to study the continuity of the reconstructed video. The results show that video frames experienced the lowest frame loss rates when transmitted using the proposed PBPD scheme. In addition, the PBPD scheme outperforms the priority-based queuing model and the best effort-based model in terms of the end-to-end delay. These results are reflected in the reduced number of starvation instants achieved by the proposed PBPD scheme at relatively higher PSNR values.