Abstract
Transparency presents an extreme challenge to stereoscopic correspondence and surface interpolation, particularly in the case of multiple transparent surfaces in the same visual direction. In this experiment we manipulate density, separation in depth, and number of transparent planes within a single experimental design, to evaluate the constraints on stereoscopic transparency. We use a novel task involving identification of patterned planes among the planes constituting the stimulus. The results show that, under these conditions, (1) subjects are able to perceive up to five transparent surfaces concurrently; (2) the transparency percept is impaired by increasing the texture density; (3) the transparency percept is initially enhanced by increasing the disparity between the surfaces; (4) the percept begins to degrade as disparity between surfaces is further increased beyond an optimal disparity, which is a function of element density. Specifically, at higher texture densities the optimal disparity shifts to smaller values. This interaction between disparity and texture density is surprising, but it can account for discrepancies in the existing literature. We are currently testing extended correlational and feature-based models of stereopsis with our stimuli. This will provide insight into our psychophysical results and a basis for quantitative evaluation of existing computational models.