Abstract
Introduction: Bingham and Lind (in press) have recently shown that Euclidean structure is perceived accurately with large perspective changes, namely a continuous 45° rotation. Their results were obtained by testing the perception of elliptical cylinders. Questions remain. First, it is unclear whether any 45° of perspective change or only a change to a 45° view of the major axis of the ellipse is important. A second question is whether performance would improve in proportion as more perspective change is provided. We investigated both questions.
Methods: A 3D cylindrical object was shown with stereo and structure-from-motion and then, a 2D ellipse appeared on the computer screen. By pressing computer keys observers adjusted the shape of the ellipse so as to match the shape of the top of the object. In Experiment 1, we compared full and half rotation conditions. In the full condition, the object was rotated by 45° to one side from a canonical view looking straight down the major axis so that a 45° view of the axis was shown. In the half condition, the object was rotated by 22.5° to either side from a canonical view so that the perspective was changed by a total of 45° but the axis was not viewed from 45°. In Experiment 2, we used different rotation amounts, 11.25°, 22.5°, 30°, 37.5° or 45°.
Results: The results of Experiment 1 showed that there was no difference between the two rotation conditions. In Experiment 2, we found that the only significant difference occurred between the three smaller perspective changes (11.25°, 22.5°, and 30°) and the two larger perspective changes (37.5° and 45°).
Conclusion: Performance does not improve in proportion as more perspective change is provided and any 45° of continuous perspective change is both sufficient and necessary to perceive metric shape accurately.