Abstract: A kind of clustering algorithm for color image segmentation is proposed, choosing HSI representation as the color segmentation space, in the study of H component, the circulation characteristics were full consideration, also the distance of two points and centers of H component is defined. Choosing the most important component of H and I component as segmentation clustering features, using fuzzy membership function depicts the clustering features, at last K-means clustering algorithm is used to segment color image. An experiments is given on bone regional segmentation, and achieve good segmentation results, experimental results shows that the results of the H and I components is superior to the results on a single component of HIS space.