Computing shape using a theory of human stereo vision