基于Gabor变换和Zernike矩的人脸特征提取方法

针对目前常用的三种人脸特征提取方法中存在的识别率低、抗噪性较弱的问题,提出一种基于Gabor变换和Zernike矩的人脸特征提取方法.该方法首先对人脸进行多分辨的Gabor变换,然后利用Zernike矩获得具有平移、尺度、旋转不变性的特征,并用线性判别分析(LDA)方法进一步进行特征选择,最后采用K最近邻分类方法进行人脸的识别.实验结果表明,在与常用的三种人脸特征提取方法的比较中,该方法具有更高的识别率和更强的抗噪性能.

Facial feature extraction method based on Gabor transform and Zernike moment

In viewing of the discriminantion feature extraction problem in face recognition, an algorithm of facial image feature extraction based on Gabor transformation and Zernike moment was proposed. This algorithm cencerned local and global facial characteristics. Firstly, the multi-resolution Gabor transformation was used for face transformaion, then the use of orthogonal Zernike moment had access to translation, sealing and rotation invariant features. Linear Discriminant Analysis (LDA) feature selection method was used for further characteristics extraction. Finally, K-nearest neighbor classification method was applied to face image recognition. The experimental results show that the proposed method is of higher recognition rate and stronger anti-noise ability.