An online weight-based clustering algorithm for faces selection

With the scale of the application of face recognition getting bigger and bigger, A thinking of selecting faces as a preprocessing has conducted on how to improve the system's efficiency. After summarized and analyzed such a thinking, this paper has discussed the pros and cons of face data clustering by using the traditional K-means clustering algorithm. In addition, this paper has further proposed an online weight-based clustering algorithm (OWCA for short) that strengthens the advantages of face data clustering, restricts the computational complexity and develops an online processing scheme. Experimental results on the ORL face library demonstrated the correctness and effectiveness of OWCA.