Managing the doubt in fuzzy clustering by means of interval-valued fuzzy sets.

In this work we study how the outliers can distort a partitional clustering process. We present a new algorithm to avoid this distortion. It is based on the minimization of a new objective functions, which is an extension of the one of the Fuzzy Clusters Means algorithm. The main novelty is the use of interval values to calculate the membership degrees of each datum to each cluster. We show the performance of our proposal over different datasets and we present its advantages in image segmentation.