Hand segmentation for hand-based biometrics in complex environments

Hand-based biometric techniques, such as the ones based on palmprint, hand vein and hand shape, is becoming more important because of their convenience and high performance. Hand segmentation is one of the most important steps in these techniques. It is a challenge task to accurately segment hand in complex environment because of the complex background, varying illuminance and other unexpected interference factors. This paper proposes a novel approach to segment hand in complex environment using color and boundary information. In the proposed approach, the hand skin color model (HSCM) is firstly constructed by using artificial neural network (ANN). Then the HSCM is used to generate a probability map (PM) and the hand is roughly segmented from the complex background by thresholding PM. After that, the hand boundary is extracted from the original image by edge detecting and voting techniques. Finally, the hand boundary is employed to cut the roughly segmented hand to get the final segmented hand. The experimental results show that the proposed approach can effectively segment hand in complex environment.