If visible light is a continuous spectrum, what's so special about RGB as primary colors?

Wikipedia goes on to say that those specific colors are close to what each kind of human cone cell is tuned to respond to, but noted that the correspondence isn’t exact. With a more modern, more precise, understanding, could we develop a better model?

And can we make devices that emit light across an arbitrary range of visible frequencies, not just one color?

The only thing special about them is that our eyes happen to have cells specialized to detect them. That’s not even a universal trait among animals; many other animals have a different set of color-detecting cells than we do. Some birds for example have a set that extends their range into the ultraviolet. Or for an extreme case you have the mantis shrimp.

Our mere 3 primary colors are pathetically limited in comparison.

They’re a bit off. The cones, especially L (red) is not most selective for red, but it is “redder” than the other two. Image of CIE1931. The “horseshoe” edge has numbers on it, corresponding to the nm in visual spectrum. The “red” cone is somewhere between the 560 and 580 range (people can have different peaks), clearly a yellow-orange range. M cone is more teal, S cone is more purple.

RGB is a color system that uses three specific primaries. With the right hue of RGB, they mix when added together to make a white. If the white is yellowish, it means e.g. the blue needs to be bluer. Your monitor uses these, I presume because of limitations of the electronic/chemical technology? (technical people, please clarify). There is no reason why you can’t have an RGBY monitor, they exist but they’re just rarer. Sharp makes a 4-primary TV, there was an old thread about that technology but I don’t remember any consensus about how it worked. Anyway, in the above image, the smaller triangle shape shows the range what a RGB monitor can show (not necessarily your monitor, which may be different, but a prototypical one). Monitors can’t use the CMY(K) model because they use light, not pigments. And of course, CRT and LCD monitors use completely different technologies to do similar things. CRT is more additive, LCD has a constant white and filters out colors.

Yes there are devices that work that way, e.g. one I’ve used: here. It’s not a monitor though, it only displays one color at a time. It works roughly by creating a bright white light, and a DLP filters out the ones you don’t need.

ETA: re: Der Trihs: even if we had UV cones/S cone was shifted, we wouldn’t see very far because our lens and macula filter out lights at that wavelength, probably as a protective measure.

ETA2: the “Formics” in the Ender’s Game film has mantis shrimp-like eyes.

Color and light SEEMS much more simple than it actually is. It has been a while since I have read in depth on this subject, but my guess is the answer to your primary question - is that it is a balancing act.

Not every color can be represented in RGB. Some TVs now have yellow pixels (I think it was samsung, but not sure). I don’t believe all versions of various color space models use the same definitions of RGB. You could imagine a situation where someone was willing to give up “accuracy” for want of a better word - for vibrance or the ability to show flesh tones more precisely.

If memory serves - part of the issue has to do with the overlapping of the response of the cones - while it is possible to light up a pixel so that it is 100% green (well not just possible, but easy) (0,255,0) - this doesn’t correspond with ONLY activating the “green” cone. It isn’t really a “green” cone at all - it just is activated much easier by green light - and gradually less so on either side of green. For many colors - unless you are on the extreme upper or lower end - you will be activating at least two of the cones.

For your last question - my guess is cost and lack of technology - currently LEDs and most fluorescent and other similar types of light sources - they only put out one frequency (or a few) - at the same time. While incandescent can be easily tuned up on down the scale - it is somewhat limited - and isn’t a pure color. Again it has been a while since I have looked in depth, but if memory serves - there is no cheap easily available source of light that puts out a single wavelength of light - and can be tuned from violet to red.

ETA: thelurkinghorror ninja’d me and had a better response - and got the manufacture of the TV right (hey at least I was right about it starting with an “S”)

Keep in mind also that what we see is somewhat an illusion. It is easy for the more technical among us to think that color should have a single - easy to use - numerical value.

That I should be able to take a sensor over an object - and get a value back.

You can do this - it is called a camera. Take a digital camera with auto-white correction off - and take a picture of a blank piece of paper outside at sunrise, noon, and the evening - as well as inside under incandescent and fluorescent light. What color do we think the paper is on all five occasions? Pure white of course. But the camera shows five different colors. The camera actually contains the more objective truth. Our eyes have changed each time - the paper REALLY is more yellowish when it looks that way on the monitor - we just don’t see it.

You could based the objective measure off the sun, but then you are being solarist (my made up word for discriminating based off what solar system you are in). There was a website somewhere I saw once with the HTML color codes (same as RGB) of what “pure white” would be based off the stars lighting.

If that were or anything like the whole story, the primary colors would be more like indigo, teal, and slightly-orangey-yellow: Spectral Response Curves of Cone Cells.

Check out the NCS (Natural Color System) color model.

Here’s a web-based navigator of its color space.

Interesting question, but doubtful if only impractical, since any material or method used to create light, will have some intrinsic limit in what wavelengths they emit or absorb. Just a WAG though.

It is even more messy than all of the above.

As thelurkinghorror points out - the senors in the eye are not specifically sensitive to red blue and green - they are all sensitive to all frequencies - but they have different shapes to the sensitivity curves. The curves njtt linked to are what effectively results when the eye processes the raw signals. The eye actually never really produces RGB - the basic response is directly processed in the eye into a luminance plus two chroma signals - that is actually quite close to the LAB system used in some image processing.

If you could find a set of three light sources that exactly matched the corners of the CIE colour gamut you could, in principle create essentially every possible perceivable colour. The curved top in the green would really require that you had two different greens, so four colours, but you would get close with three. This requires that you have monochromatic sources for your primaries - which tends to mean lasers.

If you don’t have a set of monochromatic sources, and rather your primary colour sources are a spread, you can’t get into the corners of the gamut, and that is why you end up with a triangle in the middle.

The problems we have with displays are manyfold. Plasma displays and CRT displays are all based upon phosphors - and there aren’t all that many to choose from. Displays based upon coloured LEDS will have a different range of primaries to choose from, and advances here may extend the gamut. LCD displays are partly limited by the dyes used in the pixels, but are also limited by the backlight. The desire for a cool, efficient, backlight means that in the past a cold fluorescent source was used, and now white LEDs. Both of these use phosphors to make light, and both are line emission - so you actually end up with much the same limitations in the primaries as a CRT does.

For TV and film the range of colours is limited by the reproduction technology, and the SMPTE standard is based upon what you can achieve with a CRT phosphor.

A pure black body radiator feeding a diffraction grating should get pretty close to a useful source. Then select parts of the spectrum and recombine. There are obvious limits - you are not exactly going to get a bright source - and the precision is limited by the slit width and selection widths possible. Might be a cool thing to play with as you could create close to any colour perceivable with precision.

The colors at the ends of the physiological opponent system, are normally characterized “red vs. green” and “blue vs. yellow” (think about how you can’t conceive a color that is both blue and yellow etc., thus they’re opposites). But in some similar spaces, the actual colors perceived by the brain past the photoreceptors are closer to what I might term “red-magenta vs. blue-green” and “purple vs. puke green,” respectively.

ETA: NCS is similar, but made for printing, not light. Sounds the poster who recommended that may be biased. :slight_smile:

Interesting. If it’s black body radiation, wouldn’t the color be decided on temperature? Care to elaborate a bit — I’m trying to visualize it?

Ha, well I used to work in graphics and pre-press. However, my current work has me entirely stuck in the RGB space. :wink:

I’ve never had to use NCS, but I find the idea of a human physiologically based opponent-model (diagram here) for reproduction, whether subtractive or additive, interesting.

Here’s a very long and detailed explanation of how color vision works: handprint : light and the eye