Abstract
Inaccurate judgment of metric 3D shape has been found in SFM studies and modeled as perceived relief structure (ambiguous depth scaling). However, large perspective changes (≤45°) were found to yield accurate judgments of width-to-depth aspect ratios (Lind & Bingham, 2008) and of slant (but only with non-coplanar points = bumpy surfaces) (Wang et al, 2018, 2019a, who replicated results with SFM, pure stereomotion, or both combined). Wang et al (2019b) simulated slant judgments using a stratified model that produced (1) 3D relief structure from two frame optic flow, (2) 3D similarity structure from relief structure under large perspective change, and (3) estimates of slant or aspect ratio. Polyhedrons yield non-coplanar points under planar surfaces. We now displayed polyhedrons to test judgments of both slant and aspect ratio.
We tested rectangular, hexagonal, or asymmetric pentagonal objects in SFM and stereo displays with rotations of 25° to 65° with 10° increments. Using red-cyan random texture anaglyphs, objects were presented 9 cm in front of a random texture background that was 18 cm behind the screen and viewed from 76 cm in front of the screen. Each rotation (e.g. 25°) was half to one side (12.5°) and then to the other side (12.5°). In Experiment 1, aspect ratios varied from 0.8 to 1.2 and top surface slants varied from 27° to 73° by 2° increments. In Experiment 2, aspect ratios were varied from 0.4 to 1.2 by 0.04 increments with horizontal tops. Ten participants for each shape in each experiment (60 total) judged slant in Experiment 1 by adjusting the orientation of a response line and aspect ratio in Experiment 2 by adjusting an object shaped outline.
Accuracy was poor until rotations of 35° or 45° where both judgments types became and remained accurate for larger rotations as predicted by the stratified model.