The problem of colour constancy is ill-posed. In order to extract surface reflectance accurately from the received colour signal, the visual system must rely on pre-imposed constraints based on properties of the natural world. Here we investigate the surface chromaticity distributions of 7 natural objects under 3 illuminations (D65, CWF and F), using a characterized Nikon D70 SLR camera. We find that these object surfaces exhibit intrinsic chromatic textures and provide a large number of reflectance samples on their own. The information may thereby be utilized to improve colour constancy over that achievable with artificial surfaces possessing single or limited chromaticities. By analyzing the pattern of the chromaticity distributions under changing illumination, we find that the distributions of within-surface cone contrasts for given objects form distinct signatures in cone-contrast space. These signatures transform predictably under changes in illumination. We suggest that this feature may be utilized to aid colour constancy.
Yazhu Ling, Milena Vurro, Anya Hurlbert, "Surface chromaticity distributions of natural objects under changing illumination" in Proc. IS&T CGIV 2008/MCS'08 4th European Conf. on Colour in Graphics, Imaging, and Vision 10th Int'l Symp. on Multispectral Colour Science, 2008, pp 263 - 267, https://doi.org/10.2352/CGIV.2008.4.1.art00057