Hand veins, being an inner body trait, demonstrated a reliable, secure and anti-spoofing biometric trait. Vein recognition performance highly depends on the precise detection of the vein Region-of-Interest (ROI). External influences like poor illumination, noise, blurring, complicated backgrounds, and incomplete capturing of hand region can affect the quality while capturing, causing performance degradation of wrist region segmentation that eventually leads to inaccurate ROI detection. In this paper, we propose a ROI localization method for wrist veins images. It consists of three steps: wrist region segmentation based on joint-entropy (JE) thresholding method, wrist orientation correction and ROI detection. The proposed segmentation method is compared with fifteen image segmentation methods using low quality wrist-vein images from PUT vein dataset and is evaluated by six segmentation evaluation metrics. Experiments show that the proposed method produced 99.524% accuracy that is 1.022% more accurate than the original JE thresholding method. |
