Object contour as a cue to object shape

In addition to improved distance information from height in the field, additional information about an object’s shape is provided by the projected view of the object’s upper surface and boundary contour, and by improved disparity information across a disparity discontinuity. Gillam, Flagg and Finlay (1984) investigated the role of disparity discontinuities in determining perceived slant in stimuli subject to the geometric effect. In the geometric effect a horizontal magnifier placed in front of an eye causes a frontoparallel surface to be perceived as slanted about its vertical axis. Magnification in one eye introduces conflict between binocular cues produced by the magnification and monocular cues such as perspective, with the result that the perceived slant of the surface is less than that predicted by the magnification. Gilliam et al. presented observers with a regular frontoparallel dotted grid surface. In one condition a horizontal magnifier was placed over the full field of view in one eye,

causing the whole surface to appear slanted about its vertical axis. In another condition, the magnifier was placed over just the upper half of the field of view; this caused only the upper half of the surface to appear slanted around its vertical axis and created a disparity discontinuity across the upper and lower half of the stimulus. In the full field magnification condition slant was less than that predicted from image magnification, as would be expected from the visual system combining information from the conflicting binocular and monocular cues to slant in the stimulus. However, when only the upper half of the image was magnified its slant was much closer to that predicted from the magnification (Gillam et al., 1984). This indicates that relative disparity over the disparity discontinuity provided additional information for the perception of surface slant (van Ee et al., 1999). Observers are also faster to perceive structure in random dot stereograms containing a disparity discontinuity than in full field slant stimuli, or hinge ‘open book’ stimuli containing two abutting slanted surfaces (Gillam, Chambers & Russo, 1988). This suggests that disparity discontinuities are salient image features that can aid fusion and provide relative disparity information for the estimation of surface properties. This may be beneficial as the visual system has increased sensitivity to relative disparity compared to changes in absolute disparity (Howard & Rogers, 2002).

Consider viewing a cylindrical object standing on a table surface within near visual space from above. Viewing the cylinder from above provides a projected view of its upper surface and boundary contour that are not available when the cylinder is viewed at eye height (see the ‘Lidded’ diagram in Figure 4.2). These sources of information can provide the observer with additional information with which to estimate 3-D shape. The aspect ratio of the elliptical projection of the cylinder’s upper surface at the retina can be used to determine the shape of the cylinder if the orientation, distance and height of the cylinder are known. This is because the elliptical projection of the cylinder’s upper surface changes both with the shape of the cylinder itself, and its position and orientation with respect to the observer. Information about the distance and height of the cylinder is available from both vergence and height in the field. Orientation information is provided by the disparity-defined slant of the cylinder’s upper surface and relative disparity information over the disparity discontinuity across the cylinder’s sides and upper surface.

Now consider a situation in which the cylinder is still viewed from above at the same distance and height, but now information from the upper surface of the cylinder is not available to the observer (see the ‘Contoured’ diagram in Figure 4.2). Information about the cylinder’s aspect ratio and boundary contour is still available, but in a different form, as only half the extent of the cylinder’s depth can now be seen. The observer also has the same information available from vergence and height in the field about object distance and height. However, in this situation there is no information about the slant of the cylinder’s upper surface and no relative disparity information across a disparity discontinuity. This may make it more difficult for the observer to determine the orientation and shape of the cylinder compared to when a view of the cylinder’s upper surface was available.

In summary, viewing an object from above can provide valuable information about 3- D shape that is not available when viewing an object at eye height. Eye height viewing is the typical situation in most experiments that have investigated the perception of 3-D shape from binocular visual cues. It is therefore possible that the perceptual distortions of 3-D shape demonstrated in such tasks is in part due to the unnatural viewing situation that observers are subject to. The current study focuses on understanding the role of height in the scene and a view of an object’s upper surface and boundary contour in the estimation of 3-D shape of disparity defined objects. Height in the scene can provide improved information about object distance, from which disparity can be scaled and a view of an objects upper surface and contour can provide improved information about object shape.

4.2

Methods