Perceived 3d layout in stereoscopic photographs is often distorted relative to the actual layout. The distortions include the cardboard-cutout effect, puppet-theater effect, depth shearing, depth compression and expansion, and more. Creating the correct retinal images involves the capture parameters (how the photographs were generated) and viewing parameters (how the viewer's eyes are positioned relative to the photographs). For the retinal images to be geometrically correct, the capture and viewing parameters must be compatible: the viewer's eyes must be at the centers of projection of the photographs, a constraint that is rarely satisfied. Studies in the graphics, television, and cinema literatures have investigated how capture and viewing parameters affect the retinal images and have attempted to predict perceived layout when the center-of-projection constraint is not satisfied. Those analyses were generally limited to points in the horizontal (X–Z) plane. When the stimulus extends beyond that plane, the analysis is much more complicated. Sometimes there is no geometric solution for the depth specified by the photographs. We developed a general model that predicts perceived layout from stereo photographs for all capture and viewing parameters under the assumption that the visual system does not compensate for viewing from incorrect positions. To test the model (and the no-compensation assumption), we conducted an experiment in which observers viewed stereograms from correct and incorrect viewing positions. The stereograms depicted a plane, but we will examine more complex stimuli. Observers adjusted the plane's slant until it appeared frontoparallel. Their settings were quite similar to the model's predictions, which means that viewers do not compensate for the retinal-image distortions that occur when viewing stereo photographs from incorrect viewing positions. This finding is notably different from the case of conventional photographs, where viewers compensate for incorrect viewing position. The implications for creating and viewing stereo photographs will be discussed.
NIH EY014194 and NSF Graduate Fellowship