M. Hara, T. Tokuda, T. Matsumoto, A. Matsui, and S. Clippingdale (Japan)
deformable template matching, person-independent face localization, Dirichlet process, EM algorithm
Clustering training data for person-independent face localization under expression changes often improves the performance of the deformable template matching method ([1]). One of the problems in this setting is the fact that the number of clusters is unknown. In this paper, we propose an algorithm for person-independent face localization under expression changes using clustering with a Dirichlet Process Expectation Maximum (DP-EM) algorithm ([2], [3], [4]), which directly incorporates the fact that the number of clusters is unknown. The proposed algorithm was tested against real data and compared with two conventional algorithms. One is a conventional Maximum A Posteriori Expectation Maximum (MAP-EM) algorithm ([5]) without clustering, and the other is a conventional MAP-EM algorithm with clustering where the number of clusters is known and fixed. The proposed algorithm was shown to be effective and to outperform conventional methods.
Important Links:
Go Back
