Skewed alpha-stable distribution for natural texture modeling and segmentation in contourlet domain

Abstract

Texture modeling is a very useful tool in image analysis. This model can be used in texture segmentation, denoising or texture synthesis. In this work, alpha-stable distribution has been proposed to model and segment textured images in contourlet domain. Contourlet transform’s ability to extract texture features in different scales and directions combined with alpha-stable distribution’s modeling capabilities prove to be an effective method for texture feature extraction. Kolmogorov–Smirnov distance has been used to evaluate how well the proposed distribution fits to the image in the contourlet domain. The performance of the proposed features on image segmentation has been also compared with that of features extracted using different texture analysis methods in the presence of noise. Experimental results have demonstrated the superior performance of the proposed features and their robust performance in the presence of noise.