Abstract
A picture viewed from its center of projection (CoP) generates the same retinal image as the original scene. When a picture is viewed from other locations, the retinal image specifies a different layout and shapes, but we normally do not notice these changes. On the other hand, distortions in shape are sometimes perceived when a picture is viewed from its CoP and the retinal image specifies an undistorted object. The visual mechanisms underlying such effects are unknown. We have demonstrated that perceived invariance depends on the visibility and slant of the picture surface, and consistent with our previous results does not depend on the contents of the picture. We also studied the perceived shape of objects at different positions in the picture. Observers set an ovoid to appear spherical under monocular or binocular viewing. The ovoid was presented in the center of the display or to the left and right of center. With monocular viewing through an aperture, observers always set the ovoid such that it created the retinal image of a circle. With binocular viewing, they set it such that it was a circle on the display screen, whether it was in the center the screen or not. Together with our previous results, this shows that perceived shape in pictures is determined by an adjustment to the retinal image based on measuring the local slant of the picture surface. We present a quantitative model that states how local slant measurements are made, and how they are used to adjust the retinal image. The model explains the failure of invariance with monocular viewing and the nearly veridical invariance with binocular viewing. It also explains perceived shape distortions with wide fields of view, the anamorphic effect, and the picture-in-a-picture effect.
Research supported by: NIH NRSA F32EY14514 (DV) , DOE CSGF (ARG), NIH RO1-EY12851 (MSB), AFOSR F49620 (MSB)