Abstract
Full motion video traffic is envisaged to be a major source for Internet and broadband integrated services digital networks (B-ISDN). Accurate traffic models of full motion video are needed to design networks and improve video services. Auto-regressive process (AR) proves to be a viable modeling approach of full motion video. A considerable amount of effort on AR video modeling has been reported in the recent studies which need to be thoroughly investigated. The aim of this paper is: (1) to survey a number of AR models for full motion video; (2) to classify the models according to their properties and framework; (3) to compare and contrast the models based on their attributes: residual, coding scheme, capturing scene changes, number of parameters, level of modeling, and complexity; (4) to show the ability of these models to predict accurately different aspects of network performance; (5) to give recommendations that might be helpful in determining the appropriate model for full motion video based on the target application; (6) to give direction for future work on this important modeling scheme.