Abstract
Symmetry has been found to improve 3D shape discrimination across change in viewpoint and lighting. We tested whether this benefit is due to use of symmetry as constraint for interpreting 3D information from stereo and shading. Observers performed sequential 3D shape discrimination of random, smoothly-curved symmetric objects that were rotated in depth by various amounts (0°-60°). We manipulated the 3D information provided by symmetry by varying the orientation of an object’s symmetry plane relative to the direction of viewpoint rotation. Base orientation of the symmetry plane was either horizontal or vertical, and viewpoint rotation was either horizontal or vertical. In Experiment 1, objects were presented in stereo with no shading information. We found that discrimination was better when rotation was perpendicular to the symmetry plane, which causes the symmetry plane to be slanted in depth, than when the rotation was parallel to the symmetry plane, which keeps the symmetry plane aligned with the line of sight. In Experiment 2, objects were presented monocularly with Lambertian shading, and lighting direction was varied from above-left to front-right. Performance was better for objects with a vertical symmetry plane than those with a horizontal symmetry plane with and without viewpoint rotation. In both experiments, symmetry was most advantageous when the constraint imposed by symmetry was complementary to the ambiguity from stereo or shading information.
Meeting abstract presented at VSS 2013