Palmprint Image Processing and Linear Discriminant Analysis Method

Abstract

In this paper, the method of processing and linear discriminant analysis of palmprint image is proposed. The palmprint image processing focuses on the location and segmentation which involves rotation and transition. By means of finding the two locate points about the index finger and middle finger, ring finger and little finger, the palmprint image is rotated and corrected a new coordinate system is created, which determines the region of interest (ROI). Linear discriminant analysis is a method of Gabor plus improved two-dimensional linear discriminant (Gabor+I2DLDA) which is improved 2DLDA method by integrating the Gabor wavelet representation of palm images is proposed. Traditional two-dimensional linear discriminant (2DLDA) method eliminates the column relevance of the image, while improved 2DLDA is a direct 2DLDA method which is defined in the traditional 2DLDA basis. The Gabor wavelets are used to extract palmprint features. The proposed Gabor+I2DLDA yields greater palmprint recognition accuracy while reduces the dimension. The experiment results show that location accuracy and recognition accuracy. The effectiveness of the proposed method is also verified using the PolyU palmprint databases for palmprint recognition.