STEREOSCOPIC SURFACE PERCEPTION

University Of California Berkeley, Berkeley CA

Investigators

Linked publications & trials

Abstract

The problem of visual space perception is the recovery of the location, shape, size, and orientation of objects in the scene from the light pattern reaching the eyes. The visual system uses disparities between the two retinal images to glean information about the 3D layout of the scene. However, the horizontal disparities created by a surface are an ambiguous indicator of slant and curvature. For example, such disparities are affected by slant as well as position relative to the head. Thus, the visual system must "correct" and "normalize" horizontal disparities in order to determine surface slant and curvature. Many signals could be used including vertical disparities, sensed eye position, and non-stereoscopic perspective signals Are they actually used and, if so, under what conditions? Psychophysical experiments will be conducted on slant perception with respect to various axes; vertical disparity, eye position, and perspective signals will be manipulated to determine their contributions. Experiments will also be conducted on curvature perception. There are many stereoscopic phenomena in which the perceived slant or curvature is not directly predictable from the disparities. These illusions (e.g., depth contrast, Craik-O'Brien) have been regarded as evidence that the visual system does not interpret disparity accurately and does not derive from stereopsis an accurate 3D representation of the scene. They can actually be understood. by analyzing the available signals and their statistical reliabilities. Experiments will be conducted to test this explanation's validity. Slant and curvature percepts can be modeled by linear combination of the outputs of slant and curvature estimators with different weights assigned to each estimator. The weights change appropriately with viewing situation. One can affect the weights through haptic feedback. Experiments will determine which weights are changeable via haptic feedback and how the changes generalize across conditions. From the results, one will be able to determine the level of processing at which the estimators are independent. The experimental results will be used to refine a theory of slant and curvature estimation. A better understanding of surface perception will aid development and assessment of displays used in various bioengineering applications such as remote imaging of surgery, vehicle simulators, and other virtual reality applications. A better understanding will also aid treatment of spatial distortions created when patients receive new spectacles or other optical interventions that alter image.