Buffer management for scalable video streaming

Abstract

In a video streaming scenario, to cope with bandwidth variations, buffer management and intelligent drop packets policies play a critical role in the final quality of the video received at the user side. In this context, we present a buffer management strategy implemented at the source of a video communication system. This scheme uses priority information from the H.264/SVC encoder, network information from a Bandwidth Estimation approach (BE), based on Hidden Markov Model (HMM) and monitors buffer fullness: when it exceeds a defined threshold, the selective discard strategy takes place. To get more flexibility, we employed SNR quality scalability (Medium Grain Scalability), to get more than one rate point for each enhancement layer. Low priority packets correspond to higher quality layers and are discarded first, with the aim to preserve as much as possible more relevant lower layers packets. Dependencies created by the encoding process are kept into account. We show that the strategy presented ensures that the video transmitted has the highest possible quality under the given network conditions and buffer resources.