一种基于Fourier变换的人脸识别算法

针对现有的多数人脸识别算法在单训练样本时识 别性能并不理想这一问题,提出一种基于Fourier-Mellin变换的频域不 变性以及时域不变性的特征提取算法。首先,采用图像亮度作为原始特征,将对图像亮度标 准化以改善光照变化,通过应用2D小波变换降维来管理频域不变量的复杂度；然后,为进一 步改进识别性能,根据分析的Fourier-Mellien变换(AFMT) 以及正交Fourier-Mellin矩(OFMM)法不变量的特征,将两种算法混合,进行特征提取； 最后,利用分值标准化度量频域不变量与时域不变量,并存储到一个特征向量中用于分类。 通过最近邻分类器(NNC)和相关性系 数法(CCM)进行分类和融合。通过在YALE与ORL人脸数据库上进行了大量实验的结果表明, 本文提出方法的性能要优于传统的人脸识别算法。

A face recognition algorithm based on Fourier transform

Most of the existing face recognition algorithms are not ideal when th ey are trained in a single training sample.To solve this problem, we propose a feature extraction method based on Fourier-Mellin transform in the frequency domain invariance and the moment domain invariance.Firstly,the image brightness will be normalized to improve illumina tion variation.Then,the two algorithms are mixed to extract features,according to the analytical Fourier-Mellin transform (AFMT) invariance and the orthogonal Fourier-Mellin moment (OFMM) invariance.Finally,the hybrid algorithm is classified and merged by the neares t neighbor classifier (NNC) and the correlation coefficient method (CCM).Lots of experiments on YALE and ORL face databases show that the pe rformance of our proposed method is superior to that of the traditional face r ecognition algorithms.