Characterization and source modeling of MPEG-2 VBR video source

This paper presents the results of our study on the statistical characterization of MPEG-2 VER video stream and the modeling of such source. The frame trace, marginal distribution of the frame size and the high autocorrelation existing in the continuous frames are analyzed. An ARMA process is used to model MPEG-2 video source. The different frame types of MPEG-2 video are considered together by performing a normalization transformation during modeling. The performance of the model is examined and the results show that the model is good in fitting the histogram and preserving the property of autocorrelation. We also indicate that the proper order of the ARMA model depends on the frame pattern of the MPEG source.