Abstract
Despite the prevalence of well-validated colorimetric standards, predicting the perceived gloss of surfaces from physical measurements remains challenging. The complex interactions between distal scene parameters and proximal image data make it difficult to determine mappings between physical reflectance properties and perceived gloss. Indeed, degenerate scene conditions can be constructed in which slight variations in the BRDF can drastically alter the appearance of rendered images, while in other conditions, large changes in the BRDF have hardly any visible effect. Yet, for many practical purposes (e.g., quality control), it would be extremely useful to characterize our sensitivity to changes in reflectance parameters. Here, we sought to define conditions for measuring sensitivity to reflectance changes, paving the way for standards that could serve both industry and vision researchers. Stimuli consisted of renderings of glossy objects under HDR light probe illumination. Using MLDS, we created a perceptually-uniform parameterization of the specular reflectance parameter (A) of the ABC reflectance model (Löw et al., 2012). Then, in a 2AFC task, we measured psychometric functions for gloss discrimination across a range of reflectance values when all other scene variables were held constant. JNDs increased with specular reflectance, suggesting that gloss sensitivity depends on the change in the image produced by different parameter values, even if these steps are made perceptually-uniform. To identify conditions that maximize or minimize gloss discrimination, we performed a large-scale statistical image analysis, in which we measured the magnitude of image change as a function of specular reflectance across varying distal scene variables, including shape, viewpoint, and illumination. We then measured visual gloss sensitivity for important test cases drawn from this analysis. We find that, although sensitivity varies systematically as a function of scene variables other than specular reflectance, optimal conditions for measuring and interpreting sensitivity to gloss can be determined.
Acknowledgement: EU H2020 MSC ITN No 765121 “DyViTo” and ERC Consolidator award “SHAPE” (ERC-CoG-2015-682859)