Abstract
Pictures can create very realistic impressions of 3D scenes. However, only one viewing position—the center of projection—yields the correct retinal image for the depicted 3D scene. Viewing from the wrong angle yields depth-related shearing in the image, and viewing from the wrong distance yields depth-related compression or expansion. People can compensate well for viewing from the wrong angle. It is unclear whether people can compensate for incorrect distance: Some studies found evidence for compensation, and others did not. The stimuli in these studies differed greatly. They ranged from line drawings to photographs, with and without familiar objects. Some provided low-level cues that could have allowed subjects to perform the task correctly even though they did not exhibit compensation. To determine whether compensation for incorrect distance occurs, we conducted a psychophysical experiment in which we manipulated realism and object familiarity. We also designed the experiment so that low-level cues would not allow correct performance. The stimuli consisted of pictures of two rectangular planes joined at a hinge. The planes were textured with a regular grid. We varied the hinge angle to determine the value that appeared to be 90°. The stimuli were line drawings or realistically textured computer-generated images with and without familiar shapes. Although the angle perceived as 90° differed greatly for line drawings and realistic renderings, we found no evidence that observers could compensate for viewing from the wrong distance, even with realistic renderings containing familiar objects. If they cannot compensate for incorrect distance, why do people generally not notice distortions with everyday picture viewing? We show that people naturally tend to view pictures from nearly the correct distance in part because of the lenses used by photographers.
Meeting abstract presented at VSS 2012