Abstract
For any given illumination there is a physical limit to the combinations of luminance and chromaticity that can arise by diffuse reflection alone, because reflection can only reduce the intensity of the light at each wavelength. The set of possible combinations is the theoretical object-color set of the illuminant. Naturally occurring surface reflectance is not spectrally unconstrained, so naturally occurring surfaces occupy a fraction of the object-color set. Light that does not arise from diffuse reflection of the illumination, for instance highlights or light from additional sources, can be beyond the theoretical and natural limits. We examined whether observers can detect deviations from these sets. Observers viewed patterns of 120 squares of equal size (sides of about 1 deg). All the colors but one were simulations of surfaces reflecting natural daylight. The spectral reflectances were selected at random for each trial from the surfaces of an urban scene. The simulated illuminant could be either D65 or D50. The color of the target was varied along a line crossing the border of the object color volume. The observer's task was to identify the colored square that was least likely to represent a diffusely reflecting surface under the same illumination as the other surfaces. A staircase procedure was used to estimate thresholds for six different colors for each of the illuminants. Observers could reliably identify the target as soon as it was outside or even close to the border of the natural color-set. This identification was not based on luminance or saturation alone. The ability to detect light that is unlikely to arise from diffuse reflection is probably important for color vision, because light that does not arise from reflectance should be interpreted differently to achieve color constancy.
FCT PTDC/EEA-EEL/098572/2008.