Color image segmentation using a new fuzzy clustering method

In this paper we present a fuzzy clustering method for color image segmentation. This method is based on an adaptation of a new tool for clustering called ACE (alternating cluster estimation) and a fuzzy learning by way of a neural network with a genetic training. The ACE is a new clustering model constituted of two update equations which, in contrast to classical models, are user specified and are not necessary relating to an objective function. The used neural network is composed of five layers each one, except the first, corresponds to a step of the fuzzy learning. The values optimised by the genetic training algorithm are the weights which represent the centres of membership functions characterizing linguistic terms. System input and output is respectively, colorimetric components and cluster centres. The HSV color space is utilized in order to, partly, get rid of the RGB space correlation. The experimental results obtained using the proposed method are encouraging.