Multiple resolution based palm print recognition using 2D-DWT and Kernel PCA

Palm print is a biometric pattern which possesses high discriminability due to its multiple resolution features like principle lines, wrinkles, datum points, and ridges etc. In this work, a combination of 2D-DWT and Kernel PCA have been employed for palm print based biometric recognition. Palm print images were first decomposed by 2-D Discrete Wavelet Transform and frequency band independent of multiple image resolutions was selected for dimensionality reduction. Then nonlinear mapping was applied to find the principal components for the wavelet features using kernel PCA. For image matching k-nearest neighbor's classifier has been used. The algorithm was tested on standard benchmark database (CASIA) and the results show the effectiveness of this method in terms of the Correct Recognition Rate, Equal Error Rate, and Computation Time.