Abstract
The major problem in binocular stereo vision is a well-known correspondence problem. Almost of the previous works solved it by regularization with the additional constraints such as the smoothness constraint or the ordering constraint. Therefore, they can not find the occlusion correctly and deal with the transparent surfaces, because these constraints are violated in the complicated scene contain-ing many occlusion. In this paper, we propose a method to detect binocular disparity and occlusion without using additional constraints by moving stereo cameras actively. It is based on the motion parallax obtained by the moving monocular camera. The search range of binocular disparity is restricted based on the monocular motion parallax. It uses only the Uniqueness of Disparity to find binocular disparity and occlusion. Therefore, it can find binocular disparity and occlusion correctly even for the complicated scene containing many occlusion. Furthermore, it can calculate binocular disparity very fast, because the calculation is only to count the number of disparity candidates in search range. Experimental results with complicated scene are presented to demonstrate the effectiveness of this method.
