Abstract
A powerful cue for parsing occlusion boundaries arises from relative image motion produced by an observer's self-motion (“motion parallax”), which may also utilize vestibular cues. In this situation, segmentation of adjacent surfaces is intimately associated with their perceived difference in depth. Here we investigate to what extent a depth difference generated from self-motion might aid in segmentation performance.
Stimuli appeared within a circular aperature containing a near-horizontal boundary between half-discs that were filled with 1/f (fractal) noise textures. On each trial the observer freely executed lateral head movements, which were measured using a 6-DOF electromagnetic tracking system. The textures' movements were linked to the head movement with no perceptible lag. Three relative motion conditions were compared: the two halves could move in opposite directions, in the same direction at different speeds, in sync with the head movement, or they could both move against the head movement (simulating a pair of surfaces on opposite sides of, farther than, or closer to the fixation point, respectively).
On each trial the observer reported whether the moving boundary was slightly oriented left- or right-oblique. A method of constant stimuli was used to measure a boundary orientation threshold. To examine the effect of self-motion, using comparable texture motions, we kept the head fixed with the image regions moving according to trajectories recorded during previous head-moving trials.
Orientation thresholds were smaller when the textures were moving oppositely than when moving in the same direction, and similar for the same-direction cases. Overall there was little or no difference in segmentation performance between head-free and head-fixed conditions, even though observers often reported seeing a depth difference. These results suggest that segmentation is a low level mechanism that preceeds depth perception, and that it may not benefit from vestibular input.
Supported by NSERC grant OGP0001978 to C.B.