Two-handed Gesture Tracking in the Case of Occlusion of Hands

the three-dimensional search area In this paper, we propose an efficient technique for tracking three-dimensional palm motion, which is captured by two cameras. If we perform template matching independently for each image, two points in two images do not always correspond to each other. Thus we propose to set the search area in the three-dimensional space, not in image planes, so that we can find the correct correspondences. The three-dimensional coordinates of the search area are projected on each image plane. We perform template matching at the projected point in each image. We add those two similarities and use it as a similarity at the three-dimensional coordinates. We search for the position that has the maximum similarity in the search area. When hands are crossing, we cannot track an occluded palm because the palm region is occluded by an occluding palm. Still at that situation we can track arm regions. We use the arm tracking result to solve the occlusion of hands.