The Japanese psychologist Akiyoshi Kitaoka recently tweeted a picture of some strawberries, stating that the strawberries "appear to be reddish, though the pixels are not":
It's clear that this picture does not contain any pure red, or indeed any primary colors. Images of natural scenes rarely do. But how red are the reddest pixels in the picture?
To answer this question without misleading semantic clues, I downsampled this picture in two different ways: first by averaging each local patch, and by taking the reddest color within each patch (as measured by the Euclidean distance to pure red):
This confirms the claim that most of the pixels in the original photograph aren't anywhere near pure red. However, there are a number of mauve or magenta hues suggestive of plums, grapes, or red berries. Certain pixels are close to the CSS color rosybrown.
Most of the pixels have nearly the same amount of green and blue in them. We can therefore with some justification think of the colors in the photograph as a two-dimensional space consisting of varying amounts of red and various but equal amounts of green/blue. This means that we can plot the joint distribution of channel luminosities in two dimensions:
These graphs confirm that there are no pixels with a large amount of pure red, although some of the darker hues contain slightly more red than blue or green.
Cover illustration: fruits for sale at the Municipal Market of Sāo Paolo, taken from Wikipedia.