Texture Segmentation using Complex Wavelet

Abstract

Some complex wavelet transform with directionality have been already proposed, which wecan calculate the same as the traditional wavelet transform. In this report, we propose anew supervised segmentation method based on the complex wavelet that is already proposedby us. The proposed method derives more stable results than methods using real waveletbecause the proposed complex wavelet is less shift sensitive. Although the identifcationmethod in the proposed method is a simple method using Mahalanobis distance, we achievedgood results of segmentation for some composite images from Brodatz textures.