Skin Color Weighted Disparity Competition for Hand Segmentation from Stereo Camera

This paper focuses on realtime hand segmentation from stereo camera, where the simple thresholding operation with the dense disparity map is not appropriate since it is time consuming to compute the dense disparity map, and proposes a novel method of skin color weighted disparity competition. The basic idea is that the hand disparity will be apparently different from most of the background disparities, and thus the hand pixel can be validated by verifying whether it satisfies the disparity limit of the hand more than of the background. This is why we call our method disparity competition.