Abstract
Although the role of surface-level processes has been demonstrated, models of visual interpolation emphasize contour relationships. We report research on geometric constraints governing 3D interpolation between surface patches having no visible contours. In a previous study, we demonstrated that 3D relatability acts as a cue for spatial unification in the absence of explicit edge information. Observers were asked to classify pairs of planar surface patches — specified by random dot disparities and visible through circular apertures in a fronto-parallel occluder — as aligned or misaligned. On each trial, surfaces appeared in parallel or intersecting planes with equal amounts of absolute slant around the horizontal axis. We expected this task to be facilitated when patches were perceived as connected. We found enhanced sensitivity and speed for 3D relatable patches — those joinable by a smooth, monotonic connection — versus nonrelatable patches. Here 3D relatability involves not oriented edges, but surface patch orientations computed from stereoscopic information. In a second experiment surfaces were inclined around a vertical, rather than horizontal, axis. Performance was markedly affected by slant anisotropy: both sensitivity and speed were worse for surfaces inclined around the vertical. We found nearly identical advantages of 3D relatability on performance, suggesting more isotropic unit formation effects. Results are interpreted as suggesting that, in the absence of explicit edge information, virtual lines — as extracted on the basis of orientation disparity of surface elements — constrain surface interpolation. The results suggest that 3D contour and surface interpolation may share common geometric constraints.
Supported by National Eye Institute EY13518 to PJK and a Fulbright (CIES) award to CF.