Texture synthesis using angular wavelet frames

We propose a method for texture synthesis using angular wavelet frames (AWF). The 2D AWF is constructed by frequency transformations of the prototype 1D perfect reconstruction filter bank (PRFB). The objective of using AWF is to improve the performance of existing wavelet based methods for synthesis of textures with long-range and nonlinear spatial correlation. An autoregressive (AR) model is used for each subband in the AWF decomposition. Encouraging results have been obtained by this method.