Abstract
The weighted multiresolution process model (WiMP) has been shown to combine the scale and time frequency localization properties of the wavelet representation with the self-affine characteristics of signals generated by iterated function systems (IFSs). It maintains the properties of the IFS model and resolves the instability of the inverse problem by introducing more parameters. Based on the conjecture that while IFSs are deterministic models they can synthesize natural looking textures, examples of modeling and reconstruction of real textures using WiMPs are presented. It is observed that, in the case of natural textures, the performance of the model is sensitive to the space and frequency localization of the wavelet decomposition. The deterministic model for self-affine signals considered here could be added to multiresolution techniques that lack textural information. Its localization in space makes it flexible for fitting local textures.

This publication has 4 references indexed in Scilit: