Prediction of Transmission Distortion for Wireless Video Communication: Part I: Analysis

In the wireless video communication system, transmission distortion, caused by packet transmission errors, is a non-linear time-variant function of video frame statistics, channel condition and system parameters. By modeling the transmission distortion process as output of the random video sequence and channel error processes, the system can be modeled as an equivalent non-linear time-variant system. In this paper, for the first time, we identify the governing law that describes how the transmission distortion process evolves over time, and analytically derive it as a closed-form function of frame statistics, channel condition, and system parameters through a divide-and-conquer approach. Besides deriving the transmission distortion formula, this paper also identifies two important properties of transmission distortion for the first time. The first property is that the clipping noise, produced by non-linear clipping, causes decay of propagated error. The second property is that the correlation between motion vector concealment error and propagated error is negative, and has dominant impact on transmission distortion among all other correlations. In this paper, we also identify the relationship between our result and existing models, and specify the conditions, under which those models are accurate.