Purchase this article with an account.
T. Troscianko, A. Parraga, D. Tolhurst; Is color vision good for picking fruit?. Journal of Vision 2001;1(3):61. doi: 10.1167/1.3.61.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Psychophysical measurements of human contrast sensitivity suggest that luminance is encoded in a spatially bandpass manner, whereas chrominance has a low-pass characteristic. However, the Fourier content of natural scenes usually does not reflect this different encoding — such scenes are not particularly rich in chromatic low spatial frequency information. Is this true for all such scenes? In particular, what are the spatio-chromatic properties of those scenes, such as fruit in foliage, for which color vision is thought to have evolved? We produced a set of images of natural colored objects on backgrounds of foliage using a digital camera calibrated to give relative human cone responses for each pixel. We transformed these images into red-green chrominance, and luminance, images. We measured the spectral slopes (log amplitude versus log spatial frequency) of these image pairs. If certain image types are rich in low spatial-frequency information, the chromatic slope is steeper than the luminance slope, and this situation would be in keeping with human contrast sensitivity data. We found that images of fruit and other colored objects on backgrounds of foliage are rich in low spatial-frequency chromatic content if the objects are viewed from relatively close; the effect becomes especially marked when the colored objects are at around normal grasping distance. The effect does not hold when the fruit etc is seen as a distant landscape. The effect holds under both sunny and cloudy illumination. The results of this analysis suggest that human spatio-chromatic encoding is particularly well suited to the properties of a subset of natural scenes.
This PDF is available to Subscribers Only