Abstract
The reflectance properties of a surface as a function of viewing and illumination directions can be described by the bidirectional reflectance distribution function (BRDF). Because of their high dimensionality and complexity, BRDFs are often approximated by analytical models with a small number of parameters. Our aim was to impose a perceptual metric on BRDF space by measuring discrimination thresholds, similar to the approach championed by MacAdam for color. We used the microfacet model by (Löw et al. 2012; ACM TOG), based on a modified ABC distribution (Church et al. 1989; Proc. SPIE), with 5 parameters controlling diffuse reflection, the index of refraction and the amplitude, width, and fall-off of the specular lobes. We measured discrimination thresholds in this five-dimensional space under two different natural illuminations. Separate thresholds were measured for each parameter, for increments and decrements around a single central point in the space of the parameters fitted to the MERL database. Six naïve observers were asked to detect which of four rendered blobby shapes was different in material. Adaptive staircases (QUEST) were run for each parameter and each illumination condition. We observed that observers could do the task for all 5 dimensions, i.e. all of the parameters lead to perceptually noticeable differences. Thresholds for all parameters, except the amplitude of the specular lobe, exhibited a strong degree of asymmetry between increments and decrements. Discrimination thresholds were relatively stable across different illuminations. Our results show that an approach determining visual sensitivity to small material differences is viable. Our goal is to combine the measurements of small differences to those of larger differences, as obtained by multi-dimensional scaling, to provide an overall metric of material appearance.
Meeting abstract presented at VSS 2018