Abstract
MPEG-4 encoded video is expected to account for a large portion of the traffic in future wireline and wireless networks, but there are only a few studies that evaluate networking protocols and resource management schemes with such new object-based encoded video. We explore the viability of multifractal analysis in modeling MPEG-4 video traffic. First, we prove that MPEG-4 video traffic exhibits multifractal characteristics through multifractal formalism and spectral analysis. Then we apply a multifractal wavelet model to the analysis and synthesis of MPEG-4 video traffic. This model uses the Haar wavelet transform and puts a constraint on the wavelet coefficients to guarantee positivity, which results in a fast O(N) algorithm to synthesize N-point data sets. The statistical numeric results and fit procedure show this model's flexibility and accuracy in MPEG-4 video modelling.