A Novel Method for Palmprint Recognition Based on Wavelet Transform

A novel method for palmprint recognition based on wavelet transform is presented in this paper. Because palmprint's principal lines, wrinkles and ridges have characteristics of various resolutions, a method for multi-scale wavelet transform analysis is utilized in palmprint recognition. There are three steps in the scheme, which are image preprocessing, feature extraction and recognition. Original palmprint image is aligned, segmented and enhanced firstly. Then it is decomposed into multi-scale wavelet sub-images. After the wavelet sub-images are segmented into variable blocks, the mean of each block is calculated to form a normalized vector, which corresponds to wavelet sub-image. Finally, the feature vectors of each wavelet sub-image are combined into a vector, called the palmprint feature. Feature matching is based on Euclidean distance between feature vectors and nearest neighbor distance (NND) rule. Experimental results based on PolyU palmprint database illustrate that the approach is valid in palmprint recognition