I was wondering if there is a mathematic(-like) reason why there should be 3 primary colors and why these are red, yellow and blue.
I think that what there is, instead, is a biological reason. We have cones that are keyed to those colors. In other words, they’re not intrinsically primary, they’re primary to us.
Discounting the rods, which are of little use in daytime color vision, the human eye has three types of sensor: the α-, β-, and γ-cones. For this reason, human-perceived color is naturally treated as a three-coordinate system.
There is flexibility about which three coordinates to use. Red/Green/Blue are the common choice for the additive colors of a computer monitor; Cyan/Magenta/Yellow are the common choice for the “subtractive” colors of printing. Y/I/Q (and related Y/U/V systems) were introduced for compatible color TV broadcasting and, since human brain has high spatial resolution only for the luminance coordinate, is used in systems like Jpeg and Mpeg for better compression.
Although the initial responses in human retina are from the three cone types, by the time color information gets to the optic nerve the data has been transformed into three coordinates that can very roughly be described as Yellow (or Red plus Green, or White), Red minus Green, and Blue minus Yellow.
Ignoring luminance (brightness), human-perceived color is described with two dimensions. The curved triangle in the image shows the 2-D “gamut” of chromaticities that can be perceived by a typical human. The “gamut” producible by a 3-phosphor monitor will be the triangle formed by the three phosphor colors. Note that substituting Green for the Yellow in OP’s example will enlarge the gamut significantly.
Red, yellow, and blue can be combined in various ways to create all the other colors.
More of a biological reason: humans normally have trichromatic color vision enabled by three types of cone cells, each kind sensitive to a different range of wavelengths. Therefore there is a three-dimensional space of different visible colors.
Now suppose you wish to reproduce colors by mixing pigments. You need at least three to cover a good portion of a three-dimensional space, but red, yellow and blue are not your only choices (indeed, which red, which yellow, and which blue?) The idea is that each paint absorbs a different spectrum of wavelengths of light, so different mixtures result in different colors. A popular color printing process called CMYK actually uses 4 colors: cyan, magenta, yellow, and black, and produces a decent range of colors. The first three are roughly blue, red, and yellow.
ETA ninjaed…
As I just tried to explain in the discussion of the “gamuts” of monitors, a Green with much saturation can NOT be produced by any combination of red, yellow and blue light added together … unless you add a NEGATIVE amount of red light to Yellow (or Blue plus Yellow)!
I believe the red, yellow, and blue refer to mixing paints (subtractive color), as I described in my post. I don’t think anyone would try to reproduce green light by combining red and yellow. Indeed, devices like LED screens typically have RGB pixels.
As a rule, you can create any color in that diagram septimus linked to that lies within the boundaries of the polygon created by your source colors. If you only have two source colors, those are represented by two points on the Color Diagram, and you can create any color on that line. A line connecting a red point and a yellow point doesn’t extend anywhere into the green, so you can’t make green.
Color monitors have three different colors because you can span most of the color diagram with a big triangle, but you have to have one of the points in the greenish area, or else you’ll miss a lot of the chart. (One problem of the chart is that it doesn’t represent colors “equally”. The regions of colors which aren’t distinguishable form really big ellipses (Macadam ellipses) up at the top, and much smaller elipses below. There have been various schemes to re-organize the chart into regions of equal uncertainty.
I’ve also seen attempts to create viewers that give a more faithful representation of color. One Japanese patent used something like 18 different LEDs to more faithfully span the color space. You could make colors with it that you can’t on a standard 3-color monitor.
Thanks for that. The x and y axes in that image are numbered, what are they measuring?
Just as 2-D space leads to four cardinal directions: {North, South, East, West} or {Left, Right, Up, Down} so the two dimensions of color hue lead to the four Primary colors {Red, Green, Blue, Yellow}. This thread’s emphasis on color Threeness ignores that the Four primary colors of nursery school are, for many purposes, the best set to think about!
~ ~ ~ ~ ~ ~ ~
[from the computer folklore desk] This discussion of monitor gamuts triggered my memory. Sometime in the mid 1980’s — though I can’t put a face or company name to it — I became aware of a proposal to build an LED-based display using red, green, and yellow LED’s! (Blue LEDs didn’t exist then, at least commercially.) I don’t think the project ever got off the ground.
Short answer: CIE 1931.
From the physical definition, as well as by inspecting the chart, one can see the effect of the parameters x and y, but it would be a misleading oversimplification to try to make a statement along the lines “x is red; y is green.”
Searching Wikipedia leads to a claim that Sharp Aquos has this [ETA has RGBY, at least some models], though that model would not have been around in 1980.
The Wikipedia article gives some math details but I think it’s simplest just to treat the (x, y, z) and (X, Y, Z) spaces as arbitrary. You can Google to find formulae to convert to (r,g,b) or whatever as the need arises.
The numbers around the curved portion of the almost-triangular gamut are the wavelengths of spectral (pure) colors. I’ve read that it was Sir Isaac Newton himself who was the first to think of turning the rainbow sequence into a “closed loop” by adding Purple between Red and Violet, thus forming a color wheel — is this true?
The way I’ve read about the geometrical conceptual phenomenological space of color was as a spindle, as in the Swedish Natural Color System (NCS). It may be outdated given the monitor and gamut discussions above.
Within that framework though … Hue is by polar coordinates. Chromaticness and brightness are two linear coordinatess. Together they make up a spindle. The section in that chapter also references other geometric models of color with some interesting compare/contrast.
Questions. Is hue perception by those with dichromatic vision (“color blindness”) still well described in a polar system? Would perception by creatures with greater than three hue receptors be described better by something other than polar coordinates? Or more broadly put, would the number of hue receptors a viewer possesses impact how to best model the phenomenological space? If so in what ways?
The evolution of color vision and consequently the why we’ve developed the perceptions we have are comprehensively discussed here if anyone wants a read. Some sample bits …
Although it’s a complete hijack to the thread, those who’ve not seen it before may forgive me for linking to a picture of a hamburger based on an amazing discovery by Edwin Land.
The hamburger almost looks real — it even has green lettuce! — but the picture is created with TWO colors, not three. Throughout the image the Green and Blue pixel values are always essentially identical to each other so the image is built strictly from Red and White. The human eye can reconstruct colors that aren’t really there!
When Dr. Land demonstrated this effect by projecting red and white light onto a screen, I think the effects were much more impressive than this hamburger picture.
To me, it looks like a bunch of pink/red and gray, but I could see how the lettuce bleeds into green, when contrasted against the red of the tomato. I think it’s somewhat similar to the “what color is the dress” viral post that went around about a year or two ago. The brain is very good at filling in the gaps of color and “color correcting” where applicable.
Yeah, the hamburger brown looks brown enough when my eyes adjust for the red-only filter, but the lettuce doesn’t even look like lettuce, much less green. The other colors are shades of red and white and are unsurprising.
Oh I perceive the “lettuce” as green. How much of that though is due to imposed expectation and how much due to Opponent Process effects?
Our minds are primed to see that which is expected to be seen and will perceive it even when it is not really there. Same sort of process that makes proofreading what you have typed so difficult: you expect to see the word you meant to type there and perceive its presence, read the sentence typed as having it, even when it is not there. Some of the effect may be that we expect lettuce to be green.
But there is a red green opponent process that may be occurring. It can be experienced in the “green dot illusion” here and is explained here:
Yes, I think you have to see it reflected off a screen from two projectors before the effect becomes stunning. (Note also that a background light — e.g. if you neglected to extinguish ceiling lamps when viewing the gif — will spoil the effect. Directly if a colored reflective object is in view, but even pure white will spoil unless your monitor and lightings are all producing the same white as Mean Local Sunlight.) Such demonstrations would have a variety of convincing greens and blues, with human cortex adapting chrominance locally, or as fovea flickers.
IIRC there is some evidence that, under certain circumstances, the human eye can reconstruct colors that aren’t really there— as in truly impossible— the article on impossible colors describes some experiments but does not go into technical detail. This seems to fall into the same category of optical illusions mentioned above.