1
$\begingroup$

Since cameras have RGB sensing elements per pixel, and eyes' cones similarly detect color with red, green and blue variants. The spectral sensitivities of the eyes are something like the following, and cameras similar.:

enter image description here

An element cannot alone differentiate wavelengths either side - so a blue sensing element cannot tell the difference, for example, between a blue-green and blue-violet color by itself.

Have I got something wrong or am I missing something? Thanks

$\endgroup$
3
  • $\begingroup$ Looking at the above graph, I still don't think this makes sense, but maybe: a pixel could 'appear' violet, but only as far as knowing that the amount of green is zero? $\endgroup$
    – Jodes
    Apr 24, 2019 at 10:41
  • 1
    $\begingroup$ Human color vision actually is somewhat complicated. But you'll notice that the curves in your sensitivity vs. wavelength plot all overlap. Your eyes and visual cortex effectively calculate differences and ratios of differences of the level of stimulation of different color receptor cells, and your impression of different colors is dervied from that. $\endgroup$ Apr 24, 2019 at 12:49
  • 1
    $\begingroup$ Re, "The brain interprets blue plus a lack of green as violet." Exactly! Definitely have a look at the "Opponent process" on Wikipedia. (it's short) $\endgroup$ Apr 24, 2019 at 12:56

2 Answers 2

0
$\begingroup$

To avoid serious confusion you should differentiate between green in the RGB coding, "green" as a human sensation and green as a more or less precise wavelength.

A wavelength of 500 nm is seen green, but there are infinite wavelength mixtures that are seen more or less green (and more or less "saturated"). No RGB coded device can reproduce a pure green, but only some approximation thereof. Still worse for yellow: an RGB "yellow" is far from a pure spectral yellow (about 550 nm).

If you think of the chain

  1. light source
  2. detector (photo or TV camera)
  3. display equipment
  4. human eye
  5. human brain

there are four mapping (conversion) steps and none is (none can be) faithful. A detector has three sensing elements (like retina) and for whatever luminous input it outputs three analog signals. A very poor mapping indeed.

RGB coding is a digital conversion more or less precise according to the number of bits assigned to each signal conversion.

The display equipment does its best but each element (one for R, one for G, one for B) outputs a light far from monochromatic and also different from the sensitivity curve of sensing element.

The human eye acts like the original sensor in that in the retina there are three kinds of cones, with sensitivities roughly like you sketched, but surely different from those of camera. Moreover, the retina isn't made of cones alone - there are some cell layers accomplishing a first elaboration of the signals outputted from cones.

Then via optical nerve these signals enter the brain where a most complex elaboration is done. One - only one - of the aspects of this elaboration allows for color constancy. Very simply said, this is why we attribute the same colour to an object under very different lighting conditions.

As to your interpretation of violet perception I'm not sure you're right. First, difference in sensitivities of what you call green and red cones (usually named M and L) is impressively small. Second, a monochromatic violet (say 400 nm) excites very little those cones, but M more than L (green more than red). So the retina output is significant for S (blue) cones, weak for M cones, still weaker for L ones). Maybe you're confusing with so-called "purple colours" which are not pure spectral colours but can be obtained by mixing red and blue in various proportions.

I conclude remarking that we have only scratched the surface of this complex and fascinating subject - human colour perception.

$\endgroup$
1
  • $\begingroup$ @CL22 Please note that in my last but one paragraph I had skipped a couple of words before "sensitivities". It's not that sensitivities are small (actually they are, but it's not very important) rather that their difference is small. You may find accurate curves in wikipedia under "color vision". $\endgroup$
    – Elio Fabri
    Apr 25, 2019 at 16:15
0
$\begingroup$

I think I can answer my own question:

The eyes only perceive so-called visible light down to violet wavelengths precisely because the green elements do slightly register down to violet. The brain interprets blue plus a lack of green as violet.

Cameras don't 'interpret' anything. They just copy the amount of blue and (lack of) green, allow display equipment to output the same amounts of blue and (lack of) green, and the brain then interprets it for itself as violet.

$\endgroup$
5
  • 1
    $\begingroup$ Yes that's pretty good. It is also possible to trick the eye, the same colours can be made different ways ... and your brain can't tell the difference. (a spectrometer could). $\endgroup$ Apr 24, 2019 at 13:12
  • 1
    $\begingroup$ Reproducing colors from a camera onto a screen is far more complicated. See this article for an introduction: medium.com/hipster-color-science/… $\endgroup$ Apr 24, 2019 at 13:33
  • $\begingroup$ This is good. I strongly add that you should look at the CIE color diagram en.wikipedia.org/wiki/Chromaticity#/media/… and take careful note of the distance between 470 460 nm vs. the distance from 460 to 380. By about 460 there is "no more green" and the effect that inspired your question is truly there, but in the violet, not in the blue. So you real question should sorta be "how can we distinguish color over the wide range from 460 to 380nm?" and the answer is "we can't". This is intuitively where you were ... $\endgroup$
    – Paul Young
    Apr 25, 2019 at 16:46
  • $\begingroup$ (and the same thing happens over on the "red side") $\endgroup$
    – Paul Young
    Apr 25, 2019 at 16:57
  • $\begingroup$ Actually, the reason we see violet differently than blue is becasue the RED cone sensitivity actually goes up in the violet region, not the green one. $\endgroup$
    – Rick
    Jun 8, 2020 at 18:56

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.