Research On Two Kinds Of Image Discriminant Feature Extraction Methods

Biometic Recognition Technique is the most representative and challenging research in the current biometric fields. Biometrics researches have reached a new development peak in which all kinds of new technology and methods are developed. Nowadays, image discriminant feature extraction algorithm has been focused on by many researchers. How to extract the image discriminant feature effectively becomes a research hotspot. Two kinds of image discriminant feature extraction methods are researched in this paper. More specifically, our contributions are as follows:(1) The image discriminant feature extraction methods are summarized. The nonlinear discriminant feature extraction algorithms and the discriminant feature extraction algorithms based on the frequency domain analysis are studied explicitly. Especially the nature of the two kinds of discriminant feature extraction algorithms is analyzed. Combing the two technologies, the improved feature extraction algorithms are proposed.(2) The application of the non-subsampled contourlet transform (NSCT) is studied in this paper. The nonsubsampled contourlet transform has characteristics of good multiresolution, shift-invariance and high directionality. It can give an asymptotic optimal representation of edges and contours in image. A nonlinear discriminant analysis based on the non-subsampled contourlet transform (NSCT) for palm recognition is proposed.(3) The tensor discriminant analysis is also analyzed and studied in this paper. In this paper, the tensor discriminant analysis was conducted on the palm database,AR face database and FERET face database.