基于非线性DCT鉴别分析的人脸识别

Face Recognition based on Nonlinear DCT Discriminant Analysis

有效的特征提取是人脸图象识别的关键技术。该文提出了一种非线性DCT鉴别特征提取方法并将其应用到人脸识别中。首先对人脸图像做DCT变换并选择合适的频段,然后使用广义KDCV方法来提取这些频段中的非线性鉴别特征。在FERET人脸库上的实验证明了该方法的有效性。

It s necessary to extract more effective features,which affect the classification result directly. A nonlinear DCT discriminant feature extraction approach for face recognition is proposed. Appropriate DCT frequency bands of facial images are selected,and nonlinear discriminant features are extracted from the selected bands using a generalized KDCV discriminant method. The experimental results on the FERET Database demonstrate the effectiveness of this new approach.