Simulations of adaptation and color appearance in observers with varying spectral sensitivity

Abstract

A model of adaptation and visual coding was used to simulate how color appearance might vary among individuals that differ only in their sensitivity to wavelength. Color responses to images were calculated for cone receptors with spectral sensitivities specific to the individual, and in postreceptoral mechanisms tuned to different combinations of the cones. Adaptation was assumed to normalize sensitivity within each cone and postreceptoral channel so that the average response to an ensemble of scenes equaled the mean response in channels defined for the reference observer. Image colors were then rendered from the adapted channels' outputs. The transformed images provide an illustration of the variations in color appearance that could be attributed to differences in spectral sensitivity in otherwise identical observers adapted to identical worlds, and examples of these predictions are shown for both normal variation (e.g. in lens and macular pigment) and color deficiencies (anomalous trichromacy). The simulations highlight the role that known processes of adaptation may play in compensating color appearance for variations in sensitivity both within and across observers, and provide a novel tool for visualizing the perceptual consequences of any variation in visual sensitivity including changes associated with development or disease.

Figure 1

Changes in color appearance predicted by increasing lens pigment density. (a) The original image, as seen by a reference observer assumed to have lens density characteristic of a 12-yr old. (b) The same image filtered through a lens characteristic of an average 72-yr old. (c) The filtered image as it would appear to the same visual system but with the cones adapted to the change in the average incident spectrum. Palettes below the images show the predicted changes in the Munsell array from the World Color Survey.

Figure 2

Changes in color appearance predicted by decreasing the macular pigment density. (a) the original image, as seen by a reference observer assuming a peak density of 0.5 in the central fovea. (b) The same image as seen without macular screening, characterizing the incident light in the periphery. (c) The appearance of the peripheral image assuming the cones in the periphery are adapted to the average spectrum they are exposed to. (d) The appearance of the peripheral image if only the S cones are adapted to the changes in macular screening.

Figure 3

Changes in color appearance predicted for a protanomalous observer. (a) The original image as seen by the reference normal trichromat. (b) The image as seen by a protanomalous observer assuming adaptation only in the cones to match the average spectrum. (c) The image assuming perceived contrasts are rescaled within mechanisms tuned to the individual's cardinal axes.

Figure 4

Same as Figure 3 but for a deuteranomalous observer.

Figure 5

Simulations of color appearance changes for the protanomalous observer for the case where noise precedes the contrast gain changes. Noise was added to the palette so that it was near threshold for a normal trichromat (top left panel). The remaining panels show palettes with different noise samples processed through the modeled protanomalous observer. This amplifies both the reduced signal and the noise, which is evident as random variations in the palette colors to the trichromat but should appear as largely indistinguishable variations to the anomalous observer.