Evaluation of the spatial fuzzy c-means clustering methods on medical image segmentation

In this thesis, different fuzzy c-means (FCM) clustering algorithms are examined and their performances are tested various in image segmentation applications. In this context, we firstly examined the standard FCM algorithm then we further examined two modified FCM algorithms in detail. The computer simulations of the examined methods are carried out in MATLAB environment and various image segmentations are carried out. In experimental studies, we used various images such as artifical images, real world images and medical images (Brain MR). In the results, the weakness of the standart FCM algorithm can be seen obviously. Because the standart FCM algorithm just considers the pixel?s gray level value not intersted in the spatial relationship between neighbouring pixels. This is the most important disadvantagous of the standart FCM algorithm and the modified FCM algorithms try to fix this problem. On the other hand the modified FCM and spatial FCM algorithms produced very successful results than standart FCM algorithm. This is the most important factor that both modified FCM methods considers each pixel with its neighboring pixels.Key Words: image segmentation, fuzzy c-means, spatial fuzzy c-means, medical images