Self-similarity detection in random texture via multi-scale image analysis

A dynamic scheme is presented for generating a multi-scale image associated with self-similar patterns. By blurring with small scale parameters, zero cross boundaries are extended to geometrically singular fractal patterns. Through weighted averaging with respect to scale factors, a version of conditional probability is well defined to evaluate the chasing process for unknown attractors. The imaging process, thus, is identified via the minimization of missing risk. The proposed method is verified by simulation studies