Purchase this article with an account.
Roland Fleming, Max Wolff; The intrinsic colour of transparent materials. Journal of Vision 2011;11(11):400. https://doi.org/10.1167/11.11.400.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
As light travels through a transparent material, such as bottle glass, some proportion of the light is absorbed. The absorption is often wavelength dependent, giving these materials a certain colour. The further light travels through the material, the more is absorbed, making the emerging light progressively darker and more saturated the further it travels. This poses the visual system with a colour constancy problem that has not been considered previously, and which cannot be explained by current theories of perceptual transparency. Specifically, to estimate the intrinsic spectral transmissivity of the material, the visual system must compensate for the effects of the thickness of the piece of material. To what extent do we compensate for thickness? Which image measurements do we use to estimate transmissivity? We show that this novel colour constancy problem is theoretically more constrained than standard colour constancy, because thickness cannot affect hue, and because saturation and intensity tend to be negatively correlated with one another. Using computer simulations of glass-like materials we measured the extent to which subjects compensate for thickness when judging opacity in an asymmetric matching task. Subjects matched the colour of two glass-like pebbles that differed in thickness. We find that subjects' estimates of intrinsic colour are systematically affected by thickness, showing that they cannot completely compensate for it. In a second experiment using multi-dimensional scaling, we find that thickness and opacity are distinct but interacting perceptual dimensions. The range of apparent opacities increases with thickness, much as the range of apparent lightness increases with perceived illumination (Logvinenko and Maloney, 2006). We compare the results to the predictions of several low- and mid-level image measurements. The findings suggest a novel theory of how HSV colour space can be used for separating image colours into physical causes, of which transparency is just a special case.
This PDF is available to Subscribers Only