Segmentation of Brain Images Using the FCM Clustering Algorithm and a Rough Set

ABSTRACT

The purpose of this paper is to provide a novel approach for image segmentation which combines FCM clustering with rough set theory. The image is divided into many small regions based on empirical connections between features as shown by segmented results obtained using FCM at different clustering numbers. The difference between regions is then computed using value reductions to provide weighting factors for each characteristic, and the regions similarity is assessed using the discrepancy degree-generated equivalence link. Finally, the image is segmented by combining regions that have a complete similarity-based equivalency connection. To validate this method, several brain images simultaneously segmented. As compared to the FCM method, the experimental results show that the proposed strategy will result in much lower error rates and far more precise segmentation.

FULL TEXT