Angular Bisector Fitting of Finger Valley Edge for Palmprint Localization

Aiming at the inaccurate positioning of region of interest (ROI), a method of ROI localization based on angular bisector of finger valley edge is proposed. Firstly, the edges are extracted in the YCbCr color space, i.e. tracing the palmprint contour. Then the pseudo edges and noises are removed with the ellipse model. In addition, the actual edges are smoothed and repaired with morphological manipulation, which leads to the regionalization of valley edges between fingers. On the two obtained edge regions, the two angular bisector are fitted with least square method to detect two key-points of finger valley. The proposed method cannot only remove the disturbance in the background near the finger valley, but also overcome the deviation of ROI localization resulted from the slight rotation of the palm. The experimental results confirm that the proposed method can achieve high accuracy and verification precision in mobile device environment.