The asymmetries of colour constancy as determined through illumination discrimination using tuneable LED light sources

reveals " blue " bias for colour constancy in real and simulated scenes. On calculating metamer sets for spectrally tunable LED illuminators. Are colour constancy mechanisms biased for typical illuminations? AIC 2013. Abstract The light reflected from object surfaces changes with the spectral content of the illumination. Despite these changes, the human visual system tends to keep the colours of surfaces constant, a phenomenon known as colour constancy. Colour constancy is known to be imperfect under many conditions; however, it is unknown whether the underlying mechanisms present in the retina and the cortex are optimised for the illuminations under which they have evolved, namely, natural daylights, or for particular objects. A novel method of measuring colour constancy, by illumination discrimination, is presented and explored. This method, unlike previous methods of measuring colour constancy, allows the testing of multiple, real, illuminations with arbitrary spectral content, through the application of tuneable, multi-channel LED light sources. Data from both real scenes, under real illuminations, and computer simulations are presented which support the hypothesis that the visual system maintains higher levels of colour constancy for daylight illumination changes, and in particular in the " bluer " direction, which are also the changes most frequent in nature. The low-level cone inputs for various experimental scenes are examined which challenge all traditional theories of colour constancy supporting the conclusions that higher-level mechanisms of colour constancy are biased for particular illuminations. Furthermore, real and simulated neutral (grey) surfaces are shown to affect levels of colour constancy. Moreover, the conceptual framework for discussing colour constancy with respect to emergent LED light sources is discussed. iv Acknowledgements