Quality-Oriented Video Transmission With Pipeline Forwarding

This work proposes a quality-oriented multimedia delivery framework that tackles the issue of optimizing video broadcasting and interactive video applications over packet networks with respect to both resource utilization and user perceived quality. Previous work showed that the quality of service requirements of multimedia applications can be optimally satisfied by pipeline forwarding of packets by keeping delay controlled and resource utilization high, while enabling highly scalable network devices. These properties are key in today networks to enable valuable (i.e., chargeable for) services and to avoid that the traffic increase due to broadband video either collapses existing networks or forces the deployment of high cost, cutting-edge technology to properly upgrade them. However, the current Internet is not based on such technology and its incremental introduction raises questions on how to handle video packets generated by pipeline-forwarding-unaware sources. This work proposes to use the perceptual importance of the carried video samples to determine which packets shall be transferred with pipeline forwarding-thus receiving deterministic service-and which with a traditional, e.g., best effort or differentiated service. Two new scheduling algorithms are proposed, with an extensive analysis and simulation results, which also investigate the impact of the encoding scheme. Performance bounds have been established by comparing the proposed algorithms with an exhaustive-search approach, showing that the performance is within 2 dB PSNR from the optimal solution in the worst case.