A palm print authentication system using quantized phase feature representation

摘要

This paper presents a new low cost contactless acquisition system for palm print authentication which reduces the deformations caused due to glass plates in sensors and allows the user to place the hand freely in an unconstrained fashion. Band Limited Phase Only Correlation (BLPOC) is utilized to compare two palm images. In the conventional approach using BPLOC, regions of interest (ROIs) are extracted from palm print images and stored in a database for comparison with test images; requiring more storage space as well as computation time. Therefore, instead of storing the ROI for comparison, we stored its phase angle coefficients by quantizing the image into an integer depending on the number of quantization levels used. Although, more quantization levels lead to a more accurate representation of phase angles, they come at a cost of increased feature space. The experimental results implemented on a publicly available database and on an acquired database of 175 users are very promising thus validating this new approach for palm print authentication.