Purchase this article with an account.
Franz Faul; Perceived roughness of glossy objects: The influence of Fresnel effects and correlated image statistics. Journal of Vision 2021;21(8):1. doi: https://doi.org/10.1167/jov.21.8.1.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The roughness of a shiny surface determines how sharp the reflected image of the surroundings is, and thus whether the surface appears highly glossy or more or less matte. In a matching experiment, subjects were asked to reproduce the perceived roughness of a given surface (standard) in a comparison stimulus (match), where the standard and the match could differ in both shape and illumination. To compare the effect of the reflection model on the accuracy of the settings, this was done for two different reflectance models (bidirectional reflectance
distribution function [BRDF]). The matching errors were smaller, that is, the constancy under shape and illumination changes higher, when Fresnel effects were physically correctly reproduced in the reflectance model (Fresnel-BRDF) than when this was not the case (Ward-BRDF). The subjects’ settings in the experiment can be predicted very well by two image statistics, one of which is based on the mean edge strength and the other on a local discrete cosine transform. In particular, these predictions also reflect the empirically observed advantage of the Fresnel-BRDF. These results show that the constancy of perceived roughness across context changes may depend on the BRDF used, with Fresnel effects playing a significant role. The good prediction of subjects’ settings using the two image statistics suggests that local brightness variance, which affects both image statistics, can be used as a valid cue for surface roughness.
The Gaussian distribution has shown up repeatedly in theoretical formulations of reflectance [\(\ldots\)], and it arises from certain minimal assumptions about the statistics of a surface height function. It is usually preceded by a Fresnel coefficient and geometrical attenuation factors, and often by an arbitrary constant. Since the geometric attenuation factors are typically difficult to integrate and tend to counteract the Fresnel factor anyway, we have replaced all of these coefficients with a single normalization factor that simply insures the distribution will integrate easily and predictably over the hemisphere Ward (1992, p. 268).
This PDF is available to Subscribers Only