Year

1995

Degree Name

Doctor of Philosophy

Department

Department of Electrical and Computer Engineering

Abstract

Delivery of compressed digital video signals over packet switched communications channels has a wide variety of applications. The variable bit rate transmission supported by the Asynchronous Transfer Mode (ATM) offers efficient utilisation of the network resources through statistical multiplexing but introduces new problems to real time services. For video signals the most important matter of concern is network congestion which may lead to cell loss and quality degradation.

In this thesis four packet video coding systems incorporating two-layer coding technique are proposed to overcome the problem of cell loss. The main concept of layered coding is that the layers divide the video information according to an importance criterion. The more important layers are transmitted through high priority channels and the other layers are transmitted over low priority channels. The higher priority layer provides a lower quality version of the signal at a lower average cell-rate, whilst the second layer improves the overall video quality.

The end-to-end performance of a two-layer packet video system is compared with its corresponding single layer system which uses the two-layer coding structure but all data are transmitted through the high priority channel only. The results show that the proposed two-layer systems have superior PSNR. A cost function for evaluating a packet video system and some constraints for "fair" comparison between a two-layer system and its corresponding single layer system are also developed. The cost function includes both the performance of the video compression algorithm and the cost characteristics of the communications channel. In this study the video coding standards, MPEG-I and H.261 are used for the compression algorithm and ATM-based BISDN is considered as the communications channel in all packet video systems studied.

The characterisation of the traffic generated by the encoder of a packet video system is important both in the simulation of the encoder and in the overall analysis of the packet video communications system. It is found that the Gamma density function is an appropriate model for fitting the distribution of the cells-per-frame processes generated by the encoder in each layer of a two-layer system. An Auto-Regressive Integrated Moving Average (ARIMA) model is proposed for the traffic generated by the four twolayer encoders. This model is compared with the Autoregressive order one (AR(1)) model for all proposed two-layer systems. In all cases, ARTMA model shows better goodness-of-fit than the AR(1) model.

Share

COinS