Abstract: A two-stage iterative stereo matching algorithm is proposed. Primitives used for matching are those zerocrossings from a single △2G channel that form connected edges due to their similarity in sign and orientation. The first stage is controlled by a disparity smoothness rule which substantiates Marr-Poggio's cohesiveness of the physical world. The second stage, driven by the output of the first stage, is governed by a disparity continuity rule that represents Mayhew-Frisby's figural continuity. Therefore, the algorithm yields matches that are locally and globally consistent in the sense of the two general constraints of stereo vision mentioned above.