Purchase this article with an account.
Xiaoli He, Jacob Feldman, Manish Singh; The strong influence of contour geometry in Structure from Motion (SFM). Journal of Vision 2019;19(10):198a. doi: https://doi.org/10.1167/19.10.198a.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The SFM literature focuses almost entirely on the role of image motion, ignoring contributions of contour geometry. However, our previous work (He et al., VSS, 2017, 2018) has shown that contour geometry can play a dominant role in SFM, often even overriding inconsistencies with image motion. Here we investigate the role of contour geometry by manipulating inconsistencies with image motion. Specifically, we manipulate the shape of the aperture through which the same image motion is shown. We start with a “motion region” (such as rectangle, trapezoid, ellipse) containing dot motion consistent with 3D rotation, and transform the shape of this motion region to define a smaller aperture. In many conditions, rather than looking like dots moving behind an aperture, the contour captures and defines the perceived 3D shape. In Experiment 1, we generate smaller apertures by compressing the horizontal width of the motion region. In Experiments 2 & 3, we break coaxiality (the aperture and motion region no longer share the same axis) by either translating or rotating the aperture within the motion region. The question is how much the aperture can be transformed and still determine the 3D percept. We used Method of Constant Stimuli to find the thresholds for perceiving a rotating 3D object defined by the contour shape. In Experiment 1, even with average width ratios as low as 0.2 (ellipse) or 0.6 (rectangle), the SFM percept was still dominated by contour shape. In experiment 2, the average threshold was around 0.35 (hexagon, barrel) or 0.65 (rectangle), indicating the strong effect of contour geometry on SFM percepts. These experiments document the conditions where contour geometry overrides image motion in SFM, thereby highlighting its critical, but neglected role. They also emphasize the need for mathematical models combining contour geometry and image motion in predicting SFM percepts.
This PDF is available to Subscribers Only