摘要
This paper presents a novel face recognition algorithm. To provide additional variations to training data set, even-odd decomposition is adopted, and only the even components (half-even face images) are used for further processing. To tackle with shift-variant problem,Fourier transform is applied to half-even face images. To reduce the dimension of an image,PCA (Principle Component Analysis) features are extracted from the amplitude spectrum of half-even face images. Finally, nearest neighbor classifier is employed for the task of classification. Experimental results on ORL database show that the proposed method outperforms in terms of accuracy the conventional eigenface method which applies PCA on original images and the eigenface method which uses both the original images and their mirror images as training set.
This paper presents a novel face recognition algorithm. To provide additional variations to training data set, even-odd decomposition is adopted, and only the even components （half-even face images） are used for further processing. To tackle with shift-variant problem, Fourier transform is applied to half-even face images. To reduce the dimension of an image, PCA （Principle Component Analysis） features are extracted from the amplitude spectrum of half-even face images. Finally, nearest neighbor classifier is employed for the task of classification. Experimental results on OR.L database show that the proposed method outperforms in terms of accuracy the conventional eigenface method which applies PCA on original images and the eigenface method which uses both the original images and their mirror images as training set.
