Abstract: This paper presents a double hierarchical face clustering algorithm based on c-means. By means of the algorithm the problems in t he conventional clustering algorithms, namely, (1) every datum could only be classified into one class, which made the result less accurate in clustering the data on the boundary; and, (2) the time and space complexity in fuzzy clustering are e xcessive, unfitting for high dimensional and massive data clustering have been solved. This algorithm provides a feasible clustering method for large scale face database searching, and guarantees the accuracy on increasing the speed of searching.

Key words: clustering, face recognition, c-means clustering, search in large database