Source-channel considerations for video transmission over lossy packet networks

Abstract

In recent years, the development of multimedia technologies and high-speed networking are undergoing remarkable progress. As driven by the explosive growth of Internet, various distributed multimedia applications have been introduced. It is believed that the demand for transmission of various video services over public networks will increase dramatically in the coming few years. The performance of a video communication system depends not only on the coding process but also on network environments. It is clearly impossible to look at each area in isolation. Successful transmission of digital video over networks therefore relies on the interworking between these two distinct fields. This thesis is focused on bridging video coding process and network environments for providing reliable video transmission over lossy packet networks. Various issues related to the source encoder and the transmission channel are investigated. In this thesis, we first perform a set of experiments to study the behaviors of transmitting compressed video over different network environments. As the aim of our work is to study various video services ranging from low-bandwidth to very high bandwidth, we incorporate both ISO MPEG-2 and ITU-T H.263 video standard in our experiments. Various quality of service (QoS) parameters for a video communication session like throughput, packet delay and packet loss are studied under different packing strategies and network scenario. For interaction between the source encoder and the channel, we have developed a simple receiver-driven rate control algorithm for low bit-rate video communications over the Internet. The proposed algorithm can not only control the rate but also reduce the computations at the encoding process. As loss is inevitable in packet networks, we also investigated different feedback control methods for reducing the problem of error propagation when loss occurs in the compressed video stream. An error concealment technique is also developed for the receiver to minimize the visual degradation in output video due to the data loss in the network. It is shown that a reasonable good video quality is obtained even when there is quite a high data loss in the transmission. Finally, we have implemented a remote video monitoring system in the PC environment. The system allows multiple users to monitor a single location simultaneously through the network with a frame rate up to 12 frames/s for the CIF sequence.