Abstract :
This paper presents a novel approach to the segmentation of a textured image. Initially texture segmentation necessitates the computation of various spatial dependencies, interactions and associations between the primitives of an image texture. However, regions in texture and their primitives are not always crisply defined, indeed, they can be regarded as fuzzy subsets of an image texture. Such variable nature in the underlying texture distribution makes precise object classification a difficult task, and has thus acted as stimulus to this work. The paper describes the unsupervised classification of texture based on a fuzzy clustering technique. Comparison is made to an equivalent crisp clustering algorithm in an attempt to put an order on the peculiarities and advantages of the technique, emphasising those aspects most important to the nature of the classification problem posed. It is seen that both techniques offer a high degree of classification accuracy, but more importantly our comparative studies offer the appropriateness of the techniques to specific applications
