Video adaptation for IPTV applications

Abstract

[ACCESS RESTRICTED TO THE UNIVERSITY OF MISSOURI AT REQUEST OF AUTHOR.] In recent years, with the growing popularity of inter-networking and digital video, there arises a trend of delivering television services using the Internet Protocol, (a.k.a. IPTV). Unlike in conventional television, where the signal is usually delivered over radio-frequency broadcast, cables or satellites, IPTV usually makes use of a packet-switching network, where the bandwidth is shared among multiple users. In this thesis, we focus on providing adaptive video services according to the available resource, i.e., bandwidth, latency and computation power. To address the long delay in channel switching in IPTV, a two-step switching algorithm is proposed which introduces a mid-rate bridging stream to reduce the decoder buffering delay. In conjunction with content-aware P-frame switching, our system can largely shorten the application-layer delay which causes the major latency in IPTV channel switching, and greatly improves the quality-of-service (QoS) and quality-of-experience (QoE) of IPTV. We also propose an informed transcoding scheme, where the encoder generates multiple versions of the same video content, but only sends the highest quality stream together with the motion and mode information (side information) of other lower quality streams. When the lower quality streams need to be generated, very low-complexity transcoding can be executed with the aid of the transmitted side information (up to 100 times faster than full-search transcoding). Informed transcoding achieves the best last-hop transcoding efficiency (same as full-search transcoding, better than most other transcoding schemes), and has much less overhead at the backbone network than simulcast. Such overhead is further reduced by exploiting the cross-layer correlation of the side-information.