Maximum entropy contouring and clustering for fractal attractors with application to self-similarity coding of complex texture

Based on stochastic modeling of self-similarity processes. capturing probability for not-yet-identified pattern is represented by multi-scale image. Through detecting level set and local maxima of the multi-scale image, smooth contours and finite feature pattern are estimated for fractal attractors. Estimated feature pattern is clustered within the framework of entropy maximization to design a system of reduced affine mappings with fixed points on boundary. Geometric-structural consistency of designed code is verified through computer simulation.