Fractal dimension driven bit-rate allocation strategy for video sequence coding

In this contribution a bit-rate allocation strategy for video sequence coding, based on the principle that the distortion measure, representing the information loss as perceived by a human observer, takes its minima in correspondence of chaotic regions while maxima are attained in correspondence of structured events, is proposed. In particular a fidelity criterion based on the estimated local fractal persistence is applied to drive the quantization phase of the temporal residuals in both block-based DCT and wavelet based coders. The effectiveness of the proposed criterion in improving coding efficiency, while preserving subjective quality, is demonstrated in the case of H.263 block-based DCT coder, on the basis of simulations on a set of test sequences.