Abstract: In this paper,we proposed a method for extracting the texture features in transform domain based on Dual Tree Complex Wavelet Transform（DT－CWT）,and applied it to unsupervised textured image segmentation.The image is first decomposed by one level DT－CWT.We model the modular of complex coefficients with Gamma distribution and lognormal distributrion.Texture features are extracted using Gamma distribution parameter and Lognormal distribution parameter.Then feature vectors are smoothed by the EPNSQ filter.Finally,unsupervised image segmentation results are obtained using K－means clustering based on the features.The experiment results show the effectiveness and precision of the proposed feature extraction and segmentation method.