Abstract
Rayleigh scattering causes distant objects to appear bluer than their nearer counterparts. The phenomenon is exploited by landscape painters who add a blue tint to representations of more distant objects. Using calibrated cameras (Parraga, Troscianko and Tolhurst, 2002, Current Biology. 12, 483–487) we measured the chromatic properties of near and far surfaces in natural images of mountains, hills, meadows and railway tracks. RGB values for pixels were translated into cone-activity values and in turn these were translated into values in Blue-Yellow opponent space (BY = (S−Lum/2)/(S+Lum/2)). Regressions were calculated for each category of natural scene and the mean slope was derived (ΔBY = 0.000068*meters) revealing an increase in blue pixel activity of approximately 235% over 4000 meters. In a psychophysical study the sensitivity of observers (n = 3) to these color changes was measured. The results confirmed that observers could detect changes in color caused by shifts in distance of as little as 200 meters. We conclude that while L and M cones have wavelength sensitivities optimised for the detection of fruit and for the removal of shadows (Parraga et al 2002), the relative wavelength sensitivity of the S cone results in a BY opponent system that is efficient at the discrimination of distance, even when all other cues are removed. This may be related to the benefits, on an evolutionary timescale, gained from having the ability to discriminate distances where other cues are removed, for example in savannah and desert.