Color Face Recognition Based on Curvelet Transform

In this article, a new color feature extraction algorithm and a new hybrid color space is proposed in order to further enhance the color face recognition performance. First, it is searched the best combination way of color component images of the YIQ, the YCbCr, and the HSV color spaces transformed from RGB color space. Secondly it is decomposed by applying curvelet transform to each of color component images at the proposed hybrid color space. Thirdly it is reduced by PCA the feature set. Fourthly it is used Support Vector Machines (SVM) and k-Nearest-Neighbour (kNN) classifiers for face verification. Finally the classifier outputs are fused a weighted majority voting rule. The experiments on the color FERET database and the Georgia Tech database show that the proposed Color Curvelet Face Recognition (CCFR) algorithm improves face recognition performance.