Adaptive delivery of real-time streaming video

Thesis (M.Eng.)--Massachusetts Institute of Technology, Dept. of Electrical Engineering and Computer Science, 2001.Includes bibliographical references (p. 87-92).While there is an increasing demand for streaming video applications on the Internet, various network characteristics make the deployment of these applications more challenging than traditional Internet applications like email and the Web. The applications that transmit data over the Internet must cope with the time-varying bandwidth and delay characteristics of the Internet and must be resilient to packet loss. This thesis examines these challenges and presents a system design and implementation that ameliorates some of the important problems with video streaming over the Internet. Video sequences are typically compressed in a format such as MPEG-4 to achieve bandwidth efficiency. Video compression exploits redundancy between frames to achieve higher compression. However, packet loss can be detrimental to compressed video with interdependent frames because errors potentially propagate across many frames. While the need for low latency prevents the retransmission of all lost data, we leverage the characteristics of MPEG-4 to selectively retransmit only the most important data in order to limit the propagation of errors. We quantify the effects of packet loss on the quality of MPEG-4 video, develop an analytical model to explain these effects, and present an RTP-compatible protocol-which we call SR-RTP--to adaptively deliver higher quality video in the face of packet loss. The Internet's variable bandwidth and delay make it difficult to achieve high utilization, Tcp friendliness, and a high-quality constant playout rate; a video streaming system should adapt to these changing conditions and tailor the quality of the transmitted bitstream to available bandwidth. Traditional congestion avoidance schemes such as TCP's additive-increase/multiplicative/decrease (AIMD) cause large oscillations in transmission rates that degrade the perceptual quality of the video stream. To combat bandwidth variation, we design a scheme for performing quality adaptation of layered video for a general family of congestion control algorithms called binomial congestion control and show that a combination of smooth congestion control and clever receiver-buffered quality adaptation can reduce oscillations, increase interactivity, and deliver higher quality video for a given amount of buffering. We have integrated this selective reliability and quality adaptation into a publicly available software library. Using this system as a testbed, we show that the use of selective reliability can greatly increase the quality of received video, and that the use of binomial congestion control and receiver quality adaptation allow for increased user interactivity and better video quality.by Nicholas G. Feamster.M.Eng