2D双树复小波与判别共同矢量结合的人脸认证

Fusion of 2D dual-tree complex wavelet transform and discriminativecommon vector for face verification

针对人脸认证中的小样本问题和Gabor小波特征提取的不足,提出一种有效的人脸认证算法。对预处理后的图像进行2D双树复小波变换,将每幅图像不同尺度下多个方向的小波系数幅值作为特征矢量,表征重要的局部信息;将提取的特征矢量向判别共同矢量空间投影,进一步提取具有判别能力的特征,同时进行降维;根据用户特定阈值进行认证。ORL人脸库和FERET子库上的实验结果验证了算法的有效性。

To deal with the small sample size problem and overcome disadvantage of Gabor feature extraction in face verification,an efficient algorithm based on the Discriminative Common Vector（DCV） and 2D Dual-Tree Complex Wavelet Transform（2D DT-CWT） is proposed.The 2D DT-CWT is utilized to filter the preprocessed image,and then the amplitude coefficients of different scales and different orientations are lexicographically ordered to form its feature vector.The dimensionality of 2D DT-CWT features is usually high,so features are projected into DCV subspace to reduce the dimensionality and enhance the discriminative ability.Finally,verification is accomplished according to client-specific threshold.The experimental results on ORL database and FERET subset demonstrate the effectiveness of the proposed algorithm