Abstract
A picture of a scene creates the same retinal image as the scene itself when the picture is viewed from the correct station point. Thus, it is not surprising that people perceive the scene's 3d layout accurately when viewing the picture from that point. Viewing from another position changes the retinal image such that the image is no longer consistent with the simulated scene. Yet with normal binocular viewing, there is surprisingly little effect on perceived layout. This phenomenon has led researchers to hypothesize that the visual system compensates for changes in azimuth and distance from the correct station point. How is this accomplished? We examined the role of cues from the picture's surface (paper or CRT) in the compensation. In one experiment, observers adjusted the slant of a picture until they felt they were at the correct point. The actual station point varied by −30 to 30 deg from the surface normal. When cues to the slant of the picture's surface were minimized (monocular viewing through a pinhole), observers located the correct station point reliably. With minimal information from surface cues, they presumably were unable to compensate and thus perceived the resulting distortions in the retinal image and then found the correct station point by minimizing the distortions. When cues to surface slant were maximized (binocular viewing), observers were less able to locate the correct station point, because they were now able to compensate for obliqueness of view. In a second experiment, observers adjusted the aspect ratio of a simulated ellipsoid viewed obliquely on a CRT until they perceived it as a sphere. They were better able to perform the task when cues to the CRT's surface were minimized which again reflects increased sensitivity to distortions in the retinal image when cues to surface slant are minimized. In conclusion, we found that slant cues from the surface of a picture or CRT influence the perception of simulated scenes.