Velocity constancy in natural images

MARTÍN A.; BARRAZA J.F.; ISSOLIO L.A.

Sarasota, Florida, USA

Congreso; 6th Annual Meeting of the Vision Sciences Society; 2006

Vision Sciences Society

Purpose. Previous experiments showed that for realistic artificially generated stimuli depth cues such as parallax motion, disparity, size, perspective, and texture are necessary for velocity constancy. However, the relative size produced by relative distances would be sufficient to rescale retinal velocity. An ambiguity appears because a more distant object may be also interpreted as a smaller object located at the same distance unless one knows that those are the same object that did not change its size. Natural images provide this certainty. We investigated whether size is sufficient for velocity constancy in natural images. Methods. We performed a matching speed experiment to evaluate velocity constancy in natural images. Stimuli were generated by filming from different distances an athlete walking on a platform, which could move forward or backward at variable speeds. We processed the movies to remove all the cues except the size. Results. Results show high constancy factors for all subjects (ranging from 80 to 90%) with 83% mean. Although a little bias towards retinal velocity appears in this experiment, this is smaller than those obtained in the same condition using artificial stimuli. Conclusion. The relative size produced by the relative distance between objects provides more than 80% of the information used for velocity constancy in natural images. This suggests that the visual system would trust more on size information in natural than in artificial images perhaps, because in nature objects do not change its size.