Purchase this article with an account.
Alex Lewis, Li Zhaoping; Cone tuning curves and natural color statistics. Journal of Vision 2005;5(8):268. doi: 10.1167/5.8.268.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
We investigate whether the wavelength tuning of cones gives the maximum possible information about color and lightness in natural scenes?
Humans and other old-world primates have a large overlap in the sensitivity curves of their M- and L-cones, which peak at approximately 535 and 560nm. It is often assumed that a smaller overlap would be better for general purpose vision because it would make information from the two cone types less redundant. Consequently, most theories on evolution of the cones emphasize the specific tasks of finding reddish and yellowish foods.
We compute the mutual information between cone responses and the colors in natural scenes. We find that this information would be larger if the L-cone peaked at a longer wavelength, because the correlation of M- and L-cones would be reduced and because natural scenes show most variation at long wavelengths. However, the M- and S- cones are close to their optimum positions, and little or no information would be gained if the M-cone peaked at a lower wavelength. These results remain true for color statistics for rural scenes dominated by leaves and grass, and for statistics of tropical fruit colors.
While information about colors would increase if the L-cones were sensitive to longer wavelengths, it would reduce visual acuity at high spatial frequencies, due to increased diffraction and chromatic aberration. Color discrimination performance in near-threshold conditions would also be reduced, as information is highest when the M- and L-cones are sensitive to the same part of the spectrum when signal-to-noise is low.
These results do not contradict the popular theory that the cone sensitivities are adapted to be optimal for finding fruits. However, a simpler explanation could be that the cones are designed to give maximum information about color in natural scenes across all spatial scales and illuminations, thus giving good performance in all ecological tasks.
This PDF is available to Subscribers Only