z

Wavelet-based video codecs provide a hierarchical structure for the encoded data, which can cater to a wide variety of applications such as multimedia sys-tems. The characteristics of such an encoder and its output, however, have not been well examined. In this paper, we investigate the output characteris-tics of a wavelet-based video codec and develop a composite model to capture the trac behavior of its output video data. Wavelet decomposition transforms the input video in a hierarchical struc-ture with a number of subimages at dierent resolutions and scales. The top-level wavelet in this structure contains most of the signal energy. We rst describe the characteristics of trac generated by each subimage and the ef-fect of dropping various subimages at the encoder on the signal-to-noise ratio at the receiver. We then develop an N-state Markov model to describe the trac behavior of the top wavelet. The behavior of the remaining wavelets are then obtained through estimation, based on the correlations between these subimages at the same level of resolution and those wavelets located at an immediate higher level. In our paper, a three-state Markov model is developed. The resulting trac behavior described by various statistical properties, such as moments and correlations, etc., are then utilized to validify our model