Fuzzy clustering for color recognition application to image understanding
Lotfi Khodja, L. Foulloy, Éric Benoît · Proceedings of IEEE 5th International Fuzzy Systems · 2002
Color related information is used for image analysis purposes. The FCM, the fuzzy KNN, and a proposed interpolation method based on the Delaunay triangulation are applied for the segmentation of color images. Clustering by means of triangulations is computed by linearly interpolating the membership functions. The interpolation is obtained by piece-wise approximation over a triangulation of a set of training samples. We intend to implement this new technique in fuzzy sensors performing a symbolic description of measurements. The Delaunay triangulation has proved to be suitable for clustering tasks.