关键词: 边缘检测, 向量排序, 最小支撑树, 聚类, 加权熵

Abstract: A method of color image edge detection is described.The cluster centers are generated from peaks of the color histogram.After clustering，based on the structure of color image，first，compute the Euclidean distance between neighbor pixels corresponding to the vectors in a specific region，then generate a minimal spanning tree，compute the weighted entropy and study the segmentation threshold based on the pixel vectors in this region，finally，using the vector order statistics method，sorting the gray values and gain the edge result.Experiments show that this method can get satisfied result after processing.

Key words: edge detection, vector order statistics, minimal spanning tree, clustering, weighted entropy