New Framework of HSL System Based Color Clustering Algorithm

This paper proposes a construction of a new distance in the HSL color space by a combination of the distances of HSL scalar components and by using the fuzzy cmeans framework to derive a fuzzy color clustering method. The main contribution consists in introducing two particular parameters to control the balance between the hue distance and the luminosity distance in the color Euclidean distance. The principal idea is to decrease the luminosity-distance influence and to increase the hue-distance influence when saturation values increases, and vice-versa, so that for instance, the colored pixels with similar hue, similar saturation and different luminosity values finally belong to the same cluster. These parameters are computed from the saturation values and for that reason, the proposed clustering method supplies good results also for graylevel pixels. Results are presented on some color images to illustrate the new approach and are compared to those obtained using the standard Euclidean HSL distance.