Question about Primary Colors -- Why Red-Green-Blue?

[KingGambit]

TL;DR Summary

RGP, Primary Color

Dear PF Forum,
My friend asked me this afternoon, why primary colors are Red, Green, Blue. Well, I just answered it's because our receptor cells in the retina. The L,S,M.
But something is bothering me now.
Does our brain "translate" the color from our retina?
Supposed we see two lights from red and green lamps combined, and we preceive/see/sense it's yellow.
My question is this.
1. Does our brain translate Red + Green to Yellow?

2. I googled light wavelength, and I found
https://www.britannica.com/science/color/The-visible-spectrum
Red: 650nm
Green: 550 nm
Blue: 450 nm

It seems that
(Red + Green)/2 is 600, close to Yellow (580)
(Green + Blue)/2 is 500, about Cyan (500)
(Red + Blue)/2 is 550, it's Green not Magenta (well, I think anybody who can answer my question should know that there's no Magenta. Added by Newton, right?)

My question is this:
2.
a. Does the light become yellow?
b. Does two light meet and somehow they average their frequency?
(I know that sound doesn't work that way. Two sounds merge together they don't average their frequency)

3. What about "pure" Yellow or Cyan, or Violet which our retinas don't have the cell receptor. How can our retinas perceive color which frequencies is not in our cell?

Thank you very much

 

[kuruman]

As I understand it, the eye photoreceptors are sensitive to photon energy (or frequency) not wavelength. If "pure" yellow of 590 nm hits your eye, you would see yellow. Also yellow light under water would still look yellow to you even though its wavelength in water is about 440 nm.

You may wish to read about color perception here.

 

[jbriggs444]

2.
a. Does the light become yellow?
b. Does two light meet and somehow they average their frequency?
(I know that sound doesn't work that way. Two sounds merge together they don't average their frequency)

It sounds like you have it almost within your grasp. And here's guessing/hoping the @sophiecentaur will weigh in.

Light exists in a frequency distribution. That distribution is not limited to the visible wavelengths, but we can limit our attention to that range since that is what we can sense with our eyes.

The "colors" that we can see with our eyes are not normally pure "spectral colors". They are not pure tones that exist at one particular wavelength. Instead, they are distributions. A typical human has three distinct types of cone cells on our retina. Those cones are each more sensitive to some ranges and less to others. If one takes the inputs of all three types together, one gets three numbers: so much from the "red" cones, so much from the "green" cones and so much from the "blue" cones. From this raw input (and some other contextual cues and side-influences) the visual cortex decides what color you sense from this input.

In effect, your visual cortex is guessing at a distribution of light that fits with the three numbers that you can sense.

A computer screen can reverse engineer this algorithm to present a mix of spectral colors that produces three numbers that give the illusion of seeing a chosen distribution.

 

[DaveC426913]

3. What about "pure" Yellow or Cyan, or Violet which our retinas don't have the cell receptor. How can our retinas perceive color which frequencies is not in our cell?

Because the sensitivity curve of our receptors is not tied to a single frequency. Each one has a peak, but also a broad range.

 

[KingGambit]

As I understand it, the eye photoreceptors are sensitive to photon energy (or frequency) not wavelength. If "pure" yellow of 590 nm hits your eye, you would see yellow. Also yellow light under water would still look yellow to you even though its wavelength in water is about 440 nm.

You may wish to read about color perception here.

Now, now. It would be 300000000/0.000000590 = 508 Thz , not 590nm
But, I understand your answer
Thank you sir

 

[sophiecentaur]

Summary:: RGP, Primary Color

Does our brain "translate" the color from our retina?

It's more of a 'classification' of the incoming spectrum, using a very simple combination of three values. It's done using three 'analysis' filters (in the post from @DaveC426913 above) and three sets of detectors, which is a much biologically cheaper system than using a complete spectral analysis of the light arriving at points on the retina. You could say it's a hardware solution which produces just three useful signals. More parameters are not necessary - except in some specialised animal vision.

Take any coloured source (or reflecting surface). It will not, in general be monochromatic but a whole mixture of wavelengths. You will get three values out of your sensors (the SML 'signals') Now - there are many other spectra that will produce the same SML values and the 'perceived colours' of the sources will all 'match' each other to our vision. This system could be looked upon as being very cheap and cheerful, compared with having a full spectral analysis but we evolved with a very good compromise that allows us to get by with minimal hardware. We see a colour (or one of its metemers or matches) and give it an internal 'colour name' so we recognise it again later. The names get shared between other humans by agreement but culture / language can have a bit of an influence on how different groups of humans actually appreciate their colour environment. Distinguishing between Greens is important to forest dwellers but perhaps less important for people living within the Arctic Circle. Pinks / Skin Tones are resolved accurately for barefaced creatures like humans - it's a communication medium.
You hear a lot about some groups of people giving different names to the same colour but they will all match similar colours with other colours - with varying accuracy. However, members of any race will look at colour TV and recognise the colours they see - that proves the RGB TV display system works and that the Engineers got it pretty well right.

PS. I never understand the reason for squabbles about Wavelength and Frequency of light. The Wavelength of light was measured accurately, centuries before its Frequency so we tend to use Wavelength for general conversation about optics. However, there are times when the energy of a photon is relevant and then the Frequency is used.

 

[sophiecentaur]

A computer screen can reverse engineer this algorithm to present a mix of spectral colors that produces three numbers that gives the illusion of seeing a chosen distribution.

A point of information here. The above implies monochromatic primary sources but monochromatic sources are not efficient. What you need is fairly broad band primaries because they are brighter (essential for TV watching in daylight). https://www.researchgate.net/figure/Power-density-spectra-of-the-three-primary-colors-from-a-demo-LCD-using-a-quantum-dot_fig1_276458774 shows the sort of spectra involved. The curves are broad but do not overlap significantly (unlike the analysis curves). Your sensors get a good bashing of Red Green and Blue - so there won't be any confusion!