Abstract
Abstract: :
Purpose: Colored surfaces that match under one light need not match under another. The aim of this study was to test whether normal human cones are spectrally positioned so as to minimize this metamerism and, more generally, to maximize the information about surface color available from natural scenes under different natural illuminants. Methods: Images of rural and urban scenes were obtained with a high–resolution hyperspectral imaging system (Foster et al., 2004, Visual Neurosci., 21, 331–336), which provided estimates of surface spectral reflectance over 400–720 nm at 10–nm intervals at each point in a digital representation of spatial resolution 1344 x 1024 pixels. In computer simulations, scenes were illuminated by various daylights with correlated color temperatures of 25,000 K, 6500 K, and 4000 K, and these same lights filtered by a leafy canopy. Five thousand points in each scene were randomly selected and the proportion P of metameric pairs and Shannon’s mutual information I were calculated for each pair of illuminants. The quantity I formalizes the reduction in uncertainty about the sample points under the first illuminant given knowledge about the points under the second illuminant. Mean values of P and I were obtained over 45 scenes and all combinations of illuminant changes as a function of the spectral locations of one, two, and three cone pigments (DeMarco et al., 1992, J. Opt. Soc. Am. A, 9, 1465–1476). Results: Four optimal spectral locations were identified, three of which were within about 10 nm of the mean normal cone locations. Conclusions: The spectral positions of normal human cones are almost optimal for minimizing metamerism in natural scenes under different natural illuminants and for maximizing the available information based on surface color.
Keywords: color vision • photoreceptors: visual performance • color appearance/constancy