Visibility of surface curvature in moving patterns

Abstract

We have evaluated human observers' sensitivity to surface curvature in kinetic depth patterns. Observers viewed random-dot patterns that rotated (75°) around a central vertical axis, and they discriminated between two alternative nearly spherical surfaces that differed slightly in their relative elongations along a depth axis perpendicular to the display screen. One alternative was compressed and the other was expanded relative to a sphere, thereby preserving a qualitative difference in the two shapes and minimizing the quantitative difference in their depth extensions. We found that discriminations of surface curvature are remarkably accurate. This performance persisted when the absolute size of the two ellipsoids was randomly varied, when the dynamic boundary of the rotating ellipsoids was removed, and when the perspectivity of the projective displays was randomly varied between trials. Although the random variations in size strongly affect the variations in absolute depth and the variations in perspective affect the internal-velocity field of the patterns, these random variations do not affect the discriminations of curvature. In contrast, similar cube-like surfaces, with the same depth differences but without curvature, could not be discriminated. We conclude that surface curvature is a fundamental optical property involved in the discriminations of solid shapes and that this property may be detected at the early levels of human visual processing.