Abstract
A computational approach was used to examine whether the spectral sensitivity of melanopsin, the light sensitive protein in the intrinsically photosensitive retinal ganglion cells (ipRGCs), could play a role in chromatic adaptation by providing a calibration signal to correct for daylight variability. Estimates of the signals generated with the spectral sensitivities for each cone type, short (S), medium (M) and long (L), and for melanopsin expressing cells, for a range of objects (both natural and non-natural), under a range of daylight illuminations, were computed from existing datasets. The predictive power of any individual photoreceptor type to influence chromaticity was determined, and found to be essentially non-existent. Signals derived from combinations of more than one photoreceptor type were then considered for suitability. Signals derived solely from combinations of cone-based signals were found to have inherent correlations to chromaticity, representing spurious relationships following from the fact that chromaticity is defined in this same way. These signals were thus considered to be of little value as calibration signals. Signals derived from combinations of melanopic and cone-based signals were at least as effective in predicting chromaticity as signals derived solely from cone-based signals. It was found that one combination of a melanopic signal with a luminance-like signal, when applied with appropriate weighting to the chromaticity values for an object, successfully reduces the variance between the chromaticities of that object under different illuminants. Variations on the spectral sensitivity of melanopsin were considered, by shifting the peak of the melanopsin sensitivity function, to investigate whether the recognised spectral sensitivity of melanopsin was uniquely effective in this task. The results suggest an approach to constructing a model of visual response that incorporates a melanopic input to modify L,M,S cone signals to compensate for changes in the spectrum of the ambient daylight illumination.
Meeting abstract presented at VSS 2018