Purchase this article with an account.
Phillip Marlow, Dejan Todorović, Barton Anderson; Perception of specular materials coupled to perceived 3-D shape. Journal of Vision 2014;14(10):1318. doi: https://doi.org/10.1167/14.10.1318.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
One of the fundamental problems of material perception involves understanding how the visual system identifies the physical causes of image structure. Previous research into the perception of specularity has attempted to identify particular image properties that elicit percepts of specularity. Here, we show that identical luminance gradients can generate a percept of either a matte or specular surface depending on the surface's perceived 3D geometry. Methods. In Experiment 1, we generated 1D-grating patterns and parametrically varied the steepness of the luminance gradients. We induced changes in perceived 3D shape by varying the shape of the bounding contours. One pattern contained a sinusoidal contour that was half the grating's frequency, whereas the other was a sinusoidal contour with the same frequency of the grating. Experiment 2 manipulated the shape of 3D surfaces in order to induce analogous transformations as our grating stimuli, which was used to test multiple models of the differences in specularity we observed in Experiment 1. Observers performed specular rating judgments and paired comparisons for both experiments. Results. The 3D shape differences induced by either the sinusoidal bounding contours (Experiment 1) or the 3D stimuli (Experiment 2) induced compelling differences in the perceived specularity of identical luminance gradients. The results of Experiment 2 revealed that the critical factor modulating these differences was the perceived surface curvature in the vicinity of the luminance maxima (potential specular highlights): Surfaces that appear to contain a broader range of surface normals in these neighbourhoods appeared less specular than surfaces with a narrower range of surface normals. These results were supported by Monte Carlo simulations of matte and specular surfaces in natural light fields. Conclusions. Our results demonstrate that the visual system utilizes mid-level representations of surface shape to estimate the reflectance properties of matte and specular surfaces from otherwise identical image structure.
Meeting abstract presented at VSS 2014
This PDF is available to Subscribers Only