Source modelling for packet video

Abstract

Fast packet switching or asynchronous time division (ATD) networks have been proposed for integrated-services broadband communication systems because diversified services can be offered using unified protocols. Variable-rate video-encoding techniques are usually used in these communication systems to maintain constant video quality and to take advantage of the flexible nature of the ATD network. Since the bit rate is no longer a constant, statistical models are developed to measure the characteristics of the bit-rate sequence. A first-order Markov representation is proposed to describe the probabilistic behavior of the coded video sources. Based on these models, the ATD network protocol can be designed and analyzed. A general model is proposed for the coded source rate, and several special cases of the general model for different applications are described. The related issues of video coding in ATD networks are also discussed.<>