Abstract: An improved fuzzy clustering algorithm for edge detection is proposed,in which the gray levels and the relationship between neighborhood pixels have been disposed at the same time.Based on the theory of fuzzy set,the image is mapped from gray space to a fuzzy membership matrix,and then the members of fuzzy matrix are clustered through a fuzzy clustering algorithm,and the edge of image is extracted finally.The fuzzy membership is acquired based on the theory of thermodynamics,and the edge from rough to detail can be acquired by adjusting the temperature parameter.The experiment indicates that the method is better than C-means fuzzy clustering algorithm in the extraction edge,the rapidity of computation,and the filtration of noise.