Image segmentation using a modified fuzzy C-means clustering

The current study presents an image segmentation algorithm based on modified FCM. One of the main image characteristics is the correlation between neighboring pixels. In other words, in the image segmentation, neighboring pixels are likely to belong to the same cluster. In conventional FCM, cluster assignment is only based on pixels attributes and the way they are distributed, and at the same time pixels spatial distribution and neighboring correlation aren't often taken into consideration. In other words, pixels are perceived by conventional FCM as scattered and an array is used rather than an image matrix. Other drawbacks of conventional FCM algorithm include sensitivity to small changes in intensity in homogeneous regions as well as sensitivity to noise. To put it another way, homogeneous regions in image are segmented due to shadow or small changes in intensity. We attempted to address the problems arising out of conventional FCM by investigating spatial relationship between pixels and using a multiplicative field. The results reveal the accurate function of the proposed algorithm.