This paper proposes a parallel matching algorithm for feature-based stereo vision. The algorithm is based on feature matching

and the so-called features mean zero-crossing points detected with various sizes of Laplacian of Gaussian operators. In order to obtain candidate intervals of disparity

the disparity histogram is computed all over the image. A series of processes are performed iteratively to determine all disparities. The depth pointsare interpolated to obtain a complete depth map using the local estimates of the surfacequadratic patches and edges.