A wavelet transform approach for hierarchical texture segmentation

Abstract

An algorithm that uses both the tree-structured wavelet transform and a multiscale node linking technique for hierarchical texture segmentation is developed. The tree-structured wavelet transform, which offers a nonredundant representation, is able to zoom into dominant frequency channels containing significant information of textures and can be interpreted as the decomposition of a 2-D function with the wavelet packet basis function. Based on the tree-structure wavelet representation, a fuzzy clustering algorithm is then used to implement the multiscale node linking. Numerical experiments are given to demonstrate the performance of the new algorithm.< >