The perception of surface curvature from optical motion

Area of Science:

• Vision Science
• Perception Psychology
• Computational Neuroscience

Background:

• Human observers perceive three-dimensional (3D) surface shape from optical flow.
• The visual system processes image motion to infer object structure and movement.
• Key parameters defining surface shape include curvedness and shape characteristic.

Purpose of the Study:

• To investigate how observers perceive 3D surface shape from optical flow.
• To determine which components of the optical flow field are critical for shape perception.
• To differentiate the roles of first-order and second-order motion derivatives in shape judgments.

Main Methods:

• Observers matched the perceived shape of a rotating quadric surface patch using stereoscopic displays.
• Two experiments were conducted: one with rigid and non-rigid motion, and another with masked flow fields.
• Observers controlled shape characteristic (ratio of principal curvatures) and curvedness (sum of squared principal curvatures).

Main Results:

• Observers accurately adjusted shape characteristic under both rigid and non-rigid motion conditions.
• Observers systematically overestimated curvedness, showing poor correlation with the simulated value.
• Masking first-order motion derivatives (curl, divergence, shear) had negligible effects on shape perception.

Conclusions:

• Judgments of 3D surface shape from motion are primarily driven by second-order spatial derivatives of image displacements.
• The visual system appears more sensitive to shape characteristic than to curvedness.
• These findings highlight the specific image properties utilized by the visual system for structure-from-motion perception.