Comparison Methods for Fuzzy C-Means Initialization Applied to Image Segmentation

Fuzzy C-means (FCM) is one of the most used clustering algorithms, some research seeks to achieve a better quality in the results. It is well known that an adequate selection of the initial centroids will achieve a better clustering result. In this paper, a comparison of some initialization methods applied to the FCM algorithm is made, the experimental results suggest that the K-means++ initialization method is the most suitable for image segmentation, since it produces a better initialization condition, in addition to improve convergence times.