Analysis of fuzzy clustering and a generic fuzzy rule-based image segmentation technique

Many fuzzy clustering based techniques when applied to image segmentation do not incorporate spatial relationships of the pixels, while fuzzy rule-based image segmentation techniques are generally application dependent. Also for most of these techniques, the structure of the membership functions is predefined and parameters have to either automatically or manually derived. This paper addresses some of these issues by introducing a new generic fuzzy rule based image segmentation (GFRIS) technique, which is both application independent and can incorporate the spatial relationships of the pixels as well. A qualitative comparison is presented between the segmentation results obtained using this method and the popular fuzzy cmeans (FCM) and possibilistic c-means (PCM) algorithms using an empirical discrepancy method. The results demonstrate this approach exhibits significant improvements over these popular fuzzy clustering algorithms for a wide range of differing image types.