Abstract: As the wavelet transform cannot well express curve characteristics of face image, we propose a feature extraction and face recognition method based on curvelet transform. The face image is first decomposed with curvelets.
Low frequency coefficients are compressed to reduce dimension of feature vectors, and curvelet energy features are calculated in each high frequency subband. The reduced vectors are used to represent features of the face image. Features are then classified using support vector machine. Experimental results on Orl and Yale face databases show that the proposed method is superior to wavelet methods. It is robust to varying illumination conditions, face poses and expressions.

Key words: Curvelet transform, feature extraction, Curvelet energy feature, support vector machine, wavelet transform