Abstract
Froyen et al. (ECVP2012) presented a new phenomenon involving interactions between accretion-deletion and figure-ground geometric cues. These displays contained alternating light and dark regions with random-dot textures moving horizontally at constant speed, but in opposite directions in alternating regions. When geometric figure-ground cues (convexity or symmetry) were introduced on one set of regions, subjects perceived those regions as volumes rotating in front, whereas the other regions were perceived as translating amodally behind them. This demonstrated that accreting/deleting regions, which are usually assumed to be mandatorily perceived behind, can also be perceived as self-occluding due to rotation in 3D---at least when accretion-deletion occurs on both sides of a contour. In the current project, we asked: Can geometric figure-ground cues override the standard "depth from accretion-deletion" interpretation even when accretion-deletion takes place only on one side of a contour? We used two tasks: a relative depth task (front/back), and a motion classification task (translation/rotation). In the first experiment, only one set of alternating regions contained moving texture; the other set was static. In such displays the standard accretion-deletion account would unambiguously assign farther depth to the moving regions. However, when the moving regions were convex or symmetric, they tended to be perceived as figural, and rotating in 3D (with convexity>symmetry). In the second experiment, the traditional accretion-deletion interpretation was weakened further by giving different motion directions of to the moving regions (hence weakening motion-based grouping). Our results show that the standard "depth from accretion-deletion" interpretation is easily overridden by geometric cues to figure-ground, as well as motion-based grouping across non-adjacent regions. When this happens, the accreting-deleting surface is perceived as self-occluding, hence, as rotating. Overall, the results demonstrate a rich interaction between accretion-deletion, figure-ground, and structure-from-motion that is not captured by existing models of depth from motion.
Meeting abstract presented at VSS 2013