SHAPE FROM STEREO - A SYSTEMATIC-APPROACH USING QUADRATIC SURFACES

We used quadratic shapes in several psychophysical shape-from-stereo t asks. The shapes were elegantly represented in a 2-D parameter space b y the scale-independent shape index and the scale-dependent curvedness . Using random-dot stereograms to depict the surfaces, we found that t he shape of hyperbolic surfaces is slightly more difficult to recogniz e than the shape of elliptic surfaces. We found that curvedness (and i ndirectly, scale) has little or no influence on shape recognition.