Abstract
The ability of human observers to extract the three-dimensional shape of visual objects from optic flow depends on the estimation of the characteristics of the flow. A convenient way to describe optic flow is to decompose it into translational and rotational components. Theoretical considerations [Koenderink, 1990, in Perception and Control of Self Motion Eds R Warren, A Wertheim (Hillsdale, NJ: Lawrence Erlbaum Associates)] suggest that the local orientation and shape extracted from the optic flow depend on the direction of the axis of rotation. We performed a set of experiments to evaluate the ability of human observers to determine the elevation and the azimuth of the rotation axis in structure-from-motion displays. The stimuli were pseudo-random dot patterns. The motion of the dots corresponded to the two-dimensional projections of opaque spherical, elliptical, or hyperbolic surface patches. A matching procedure was used. The results show that the estimation of the elevation is more precise and reliable than the azimuth of the rotation axis and does not depend on differences in stimulus shape or on the azimuth values. The best estimates were obtained for vertical and horizontal axes of rotation. The results are discussed in relation to the role of rotation axis evaluation for the extraction of the local shape of the three-dimensional objects, as well as to the possible mechanisms for the determination of this global characteristic of the optic flow.
Get full access to this article
View all access options for this article.
