In conventional stereo matching, it is assumed that the camera is comprised of a thin lens and that the parallel axis constraints are satisfied; that is, the focal rays of the two cameras are parallel and perpendicular to the stereo baseline. Consequently, the two image planes are coplanar and the scan line is parallel to the baseline or epipolar line.
It is also assumed that the pixels corresponding to each point in the scene have the same brightness. If these constraints are satisfied, the search area for correspondence is restricted to a 1-D line and the matching process is accelerated significantly.
Finally, 3-D information can be computed by binocular disparity and triangulation. It can be shown that the relative depth is inversely proportional to disparity [9].