A New Texture Segmentation Method Based on the Fuzzy C-Mean Algorithm and Statistical Features
Mounir Sayadi, Lotfi Tlig, Farhat Fnaiech · 2007
The segmentation of textured images is a fundamental problem in image processing and pattern recognition, which has been addressed for different applications with several methods. In this paper, the problem of textured image segmentation with an unsupervised method is addressed. To overcome the difficulty of the segmentation of images involving complex random texture patterns, we propose a cascade clustering method combining statistical features and the standard Fuzzy C-Means clustering algorithm. Instead of using the gray level value of a given pixel, a feature vector is extracted from a sliding window centered on the pixel. The Fuzzy C-means algorithm is used to cluster the obtained feature vectors into several classes corresponding to the different regions of the multi-textured image. Experimental results show that the proposed method is a better alternative to the pixel-based FCM segmentation method.