Stochastic-computational approach to self-similarity detection in random image fields

摘要

We present an integrated stochastic-computational scheme for detecting self-similarity in random image fields. By modeling imaging as a Brownian motion in a successively reduced domain, capture probability is induced on the image plane. Attractordistribution is simultaneously identified with fixed points corresponding to mapping sequences generated by imaging. The computational structure of local maxima of capture probability is extracted through invariance and observability analysis to matchobserved attractors with a preassigned mapping dictionary. Proposed scheme was implemented as digital algorithm and verified through simulation.