A palmprint based authentication system that can work with a popular webcam in non-contact acquisition mode is potentially a good choice for biometric applications.However,this camera based imaging acquisition mode causes the difficulty for the location of palmprint due to the unstable palm position and variable illumination condition and effects the extraction of palm region of interest(ROI).In particular,changes in illumination of the system effect its performance heavily. The process of extract palm ROI has been discussed in different papers, but hardly does very well under variable light conditions and pose changes.In this paper,we propose a robust approach for localizing the palm and extracting the ROI based on real-time region learning.A dynamical region is learned to binarize the image and get the hand contour to extract the palm ROI. In a database of 1000 video clips of hand under different illumination and poses,the accurate extraction rate reaches 92%.
Meng Yan, Dongmei Sun, Shouguo Zhao
Journal name not available for this finding