Local stereo cues for estimating coplanar surface alignment

Abstract

Previous reports indicate that there is a special sense involved in coplanar alignment of isolated surface patches (Taroyan et al., 2000). There have been no quantitative tests of this claim. We tested it by comparing stereo-slant discrimination thresholds for a test surface patch with a coplanar or a non-coplanar reference surface present. Thresholds were lower with the coplanar than with the non-coplanar reference surface. This result suggests that a common surface could be estimated from cues derived from local surface patches. Backus et al. (1999) proposed a retinal cue for coplanar judgments that is independent of distance and azimuth: HSR/VSR^2 which is constant on a plane. However, non-coplanar surfaces will have the same ratio if they intersect at a point on the interocular axis. We investigated whether coplanar judgments worsened when the ratio cue was made ambiguous. In each trial, two surfaces were presented. The reference surface had a fixed slant and distance. A test surface from two surface families was presented at different distances. One family had the same HSR/VSR^2 as a reference surface, but different slants. The other had the same slant as a reference surface, but different values of HSR/VSR^2. Observers judged whether the test surface was closer or farther away than the extended surface of the reference (i.e., coplanar). If observers relied solely upon HSR/VSR^2 to make coplanar judgments, the constant-ratio condition should have higher thresholds than the constant-slant condition. The thresholds for the two stimulus configurations were not significantly different, which suggests that the ratio is not used to judge coplanar alignment, or that the cue is no more informative than other available cues.

Footnotes

 Supported by NIH grant EYO-08882