September 2011
Volume 11, Issue 11
Free
Vision Sciences Society Annual Meeting Abstract  |   September 2011
Near-optimal tuning of trichromatic vision for constant surface identification in natural scenes
Author Affiliations
  • David H. Foster
    School of Electrical and Electronic Engineering, University of Manchester, Manchester, UK
  • Sérgio M. C. Nascimento
    Centre of Physics, Gualtar Campus, University of Minho, Braga, Portugal
  • Kinjiro Amano
    School of Electrical and Electronic Engineering, University of Manchester, Manchester, UK
  • Iván Marín-Franch
    Indiana University School of Optometry, Bloomington, Indiana, USA
    Department of Optometry and Visual Science, City University London, London, UK
Journal of Vision September 2011, Vol.11, 357. doi:https://doi.org/10.1167/11.11.357
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      David H. Foster, Sérgio M. C. Nascimento, Kinjiro Amano, Iván Marín-Franch; Near-optimal tuning of trichromatic vision for constant surface identification in natural scenes. Journal of Vision 2011;11(11):357. https://doi.org/10.1167/11.11.357.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

There have been several partial attempts to explain the number and spectral locations of the cone photoreceptor pigments of the normal human eye; e.g. that specifically the medium- and long-wave pigments of trichromatic primates are optimal for discriminating fresh leaves from mature foliage. The aim of this study was to test a more general hypothesis, namely, that long-, medium-, and short-wave pigments are optimal for identifying objects within the natural world despite changes in the color of the illumination. To determine the theoretical limits on identification performance, computer simulations were performed in which the number of cones, pigment spectral locations (on a log-wavelength axis), and post-receptoral interactions were allowed to vary. Stimuli were generated from high-resolution hyperspectral images of 50 close-up and distant images of natural scenes. Of these scenes, 29 were classed as predominantly vegetated and 21 predominantly nonvegetated. From each scene, 1000 points were chosen randomly and the cone signals at each point were calculated for two illuminants selected randomly from combinations of direct sunlight and blue skylight and these same lights filtered by a leafy canopy. Allowance was made for absorption in the ocular media. The information retrieved, derived from an estimated probability mass function for identification performance, was calculated for one, two, three, and four cone pigments. The average information retrieved increased with increasing pigment number, but it approached an asymptote with three pigments. Optimal spectral locations differed according to whether the scenes were predominantly vegetated or nonvegetated, but, for both types, spectral locations were generally shifted towards longer wavelengths with respect to their normal positions, although little additional information was gained by the shift. The number and tuning of normal cone pigments seem to be nearly optimal for surface identification in the natural world.

EPSRC EP/F023669/1, FCT PTDC/SAU-BEB/108552/2008. 
×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×