Quality of Service using Scalable Video Coding and Partial Buffer Sharing

The last-mile has limited bandwidth and forms a bottleneck link for video applications. Current solutions fail to offer sufficient Quality of Service (QoS) in this last-mile because they allow packet loss randomly in a stream, while packet loss, depending on which part or layer of the video, has strong impact on the final quality. Our paper proposes to concentrate the packet loss in the least important layers of a video by combining Scalable Video Coding which offers graceful degradation of quality when removing video layers, with Partial Buffer Sharing which drops video layers proportional to their importance. Simulations reveal a fundamental issue with SVC and unequal error protection. The different scalable layers do not alternate sufficiently in time, causing the buffer to drop more important layers at times. The paper analyses and solves this problem, and demonstrates it can concentrate the packet loss in the least important video layers.