What is it about?
Color helps us recognize things in the world around us, but color depends on illumination, for example, light from the setting sun gives scenes a yellowish cast and light from the polar sky a bluish cast. It is shown here by computational methods that natural variations in daylight -- shifts in patterns of light and shade rather than in overall color -- severely limit the number of surfaces we can identify by color.
Featured Image
Photo by Peter Keller on Unsplash
Why is it important?
Research in this area has traditionally focused on our constant perception of surface color under changes in illumination color, so-called “color constancy”. Yet most variations in natural light are not in color but in the pattern of light on scenes. Because of these variations, the number of surfaces identifiable over an interval is at least an order of magnitude smaller than with overall changes in illumination color.
Read the Original
This page is a summary of: Fluctuating environmental light limits number of surfaces visually recognizable by colour, Scientific Reports, January 2021, Springer Science + Business Media,
DOI: 10.1038/s41598-020-80591-9.
You can read the full text:
Resources
Hyperspectral radiance image pairs
This collection contains pairs of successive hyperspectral radiance images of 18 natural scenes. Each image is stored in a separate dataset labelled by scene name, approximate acquisition time, and image type. Intervals between images in each pair vary from 1 to 15 minutes.
Software for estimating differential entropy and mutual information for multivariate data
Numerical estimators of differential entropy and mutual information can be slow to converge as sample size increases. The offset Kozachenko–Leonenko (KLo) method implements an offset version of the Kozachenko–Leonenko estimator that can markedly improve convergence. See https://doi.org/10.1017/exp.2022.14.
Contributors
The following have contributed to this page
