Abstract
Disparity for a single horizontal line is ambiguous. It can originate either from slant, which magnifies one image, or from occlusion by a nearer surface, which truncates the image more for the eye on the side of the surface. Even without an explicit surface present, disambiguating information can exist in the pattern of disparities among a set of parallel horizontal lines if they vary in length. Slant produces magnification proportional to width so that lines of different initial length would also have different additional lengths in the magnified image. Occlusion will simply truncate each line by a constant amount in one eye regardless of its length. Gillam et al (ECVP, 2001) found a strong subjective occluding contour along the aligned side of a set of horizontal lines of different lengths when a constant length was added for the eye on the side of the edge alignment (valid occlusion stimulus). Here we compare perceived slant for uniocular truncation and magnification for sets of horizontal lines, measuring slant for a long and a short line within each set. Both disparity directions were used, consistent with valid or invalid occlusion or with positive or negative slant. Method. A stereo probe measured perceived depth at each end of the designated line from which its slant was calculated. Results. For valid occlusion conditions the lines were perceived as approximately flat and at the depth predicted from occlusion considerations. For the invalid occlusion conditions lines were seen at different slants. For the magnification conditions uniform slant was seen, closely in accordance with prediction from the magnification present. Conclusions. Despite the complete ambiguity of each line with respect to slant/occlusion, subjects were able to perceive a line as slanted or not with considerable accuracy and precision based on a comparison of disparities across the set of lines. This reveals a novel global stereoscopic ability overriding local disparity information.