Improving network video quality with delay cognizant video coding

Abstract

Delay critical network video like videoconferencing requires bandwidth reservations higher than the long-term average rate in order to accommodate its bursty traffic output. While prior research contributed to the estimation of the required bandwidth, which is often known as the effective bandwidth, few if any of the previous work attempted to make use of the rate gap (residual bandwidth) between the average and effective rates. In this paper, we presented an application of delay cognizant video coding (DCVC) to improve the quality of network video by using this residual bandwidth. DCVC generates two compressed video flows with differential delay requirements, one delay-critical flow at the average rate to establish an initial presentation and the second delay-relaxed flow at the residual rate to improve its quality. At no additional cost to end users, significant quality improvement can be achieved and it is demonstrated through both objective and subjective measures of a simulated network video connection.