Color wheels are wrong? How color vision actually works

http://blog.asmartbear.com/color-wheels.html?

by akkartik • 15 years ago

225 54 15 years ago

blog.asmartbear.com

diiq2• 15 years ago

I'm happy to see someone investigating color, and the way we perceive it. So many color theory books are more mystical than scientific; it's wonderful to see a physiological model.

I suspect that we continue to teach the RGB, RBY, and CYMK color wheels because they behave correctly with respect to the behavior of particular physical media. Yellow-looking paint and blue-looking paint do make green-looking paint. Red-looking light and green-looking light do make yellow-looking light.

The (physiologically accurate?) four-primary-color model proposed here does not correctly predict the world: yellow and blue act as additive color opposites; red and green act as subtractive color opposites. No one set of pigments or lights will ever consistently behave the way this 4 color wheel predicts.

We need multiple systems because colored objects mix by multiple mechanisms.

sedachv• 15 years ago

It's not really a new thing. Rudolf Arnheim's Art and Visual Perception (published in 1974, widely read in art school from what I understand) had a whole chapter about exactly the same material as in that blog post. But before that it had a chapter on the psychology of color, and why the color wheel is in fact useful.

diiq2• 15 years ago

Right! I used "proposed" to mean "argued for" rather than "set forth for the first time".

Arnheim is kind enough to distinguish between "generative complements" and what he calls "fundamental complements", (though I would prefer they be called "perceptual"). He warns specifically against confusing perceptual color and color-in-the-world. By making that distinction, Arnheim validates BOTH systems, rather than attempting to invalidate three-primary systems (as this article appears to do).

The article calls RGB and RBY color "wrong" and unable to "stand up to even minor scrutiny".

Arnheim calls them "generative".

I claim their generative nature (that they inform how to create color) is why they are taught in schools. I'm not denying the four-color idea; I'm just saying it is based on a different definition of color, and different intended use of color.

sfphotoarts3• 15 years ago

while the article looks like a mashup of text-books and wikipedia (not to mention the self-aggrandization) and is interesting, this looks like SEO spam, one of these sites that sells you some book on how to get rich quick or how to start your startup.

I think I'll stick with my Johannes Itten.

michael_dorfman• 15 years ago

You talk about what the article "looks like", but have you actually read it?

I found it to be a well-written, well-researched introduction to some of the problems in color theory. If that's what you mean by "a mashup", then I guess all secondary literature is.

As for the site, asmartbear is no stranger to these parts, and Jason's posts are well-regarded.

sfphotoarts3• 15 years ago

possibly my British grammar is stumping you, but yes, to read something you have to look at it.

This article might appear well researched here, but this is the hello-world equivalent on an art forum.

Maybe if Jason believes he has something meaningful to contribute to color theory, he should consider updating the Wikipedia page rather than promoting his own self interest by advertising his book & blog on HN.

CodeMage• 15 years ago

Do you have anything to back that claim? Even with Adblock Plus disabled, I don't see anything that could be descriped as "SEO spam".

wccrawford• 15 years ago

The budding designer in me loves these articles. While I can intellectually understand the color wheel(s), I'm still having issues grasping it emotionally and creatively.

lukev• 15 years ago

I wish I could upvote this twice. I hadn't realized that our actual perception of color was so filtered from the physical spectrum.

I find it incredibly fascinating, from a philosophical perspective, that there are literally color combinations (i.e, greenish-red) that, despite lying within the physical spectrum to which our eyes are sensitive, we cannot see, cannot even imagine.

NonMint• 15 years ago

The thing that will really blow your mind is are there things that are red, or a group of things that have a common appearance that our minds just label "red." And what if what if the signals that reach my brain when I view "red" are shifted one to the left on the color wheel? The gradual perception of color change would be perceived but we would be looking at different colors even though we both agree that it is "Red."

apl2• 15 years ago

  > The thing that will really blow your mind is [...]

Only if "you" refers to a stoned college kid. The realization that qualia -- such as the redness of red -- are fundamentally subjective is, quite frankly, an old hat.

sambeau• 15 years ago

The checkerboard illusion at the end amazes me every time I see it.

joeld42• 15 years ago

This article tries to sensationalize and obfuscate something that is pretty simple in reality.

RGB are the primaries in light. By mixing these three colors you can create any color the human vision system can perceive (yes, because of tristimulus).

When white light hits a material, some of those RGB wavelengths are absorbed (subtracted). RGB - GB(Cyan) = Red, RGB - RG(Yellow) = Blue, RGB - RB(Magenta) = Green. Thus, Cyan, Magenta and Yellow act as primaries (colors from which you can construct any visible color) for materials that absorb, not emit light.

The forth primary in print media, Black, is simply added because inks are imperfect and a mix of CMY inks yields a kind of murky black (plus it wastes ink).

It's a historical simplification to say that "red, yellow and blue" are the primaries for paint, the red works like a magenta, the blue works like a cyan. They sell Cyan/Yellow/Magenta primaries for oil/acrylics and if you paint with these you'll get a brighter, wider palette but you can get there with a traditional palette by adding whites and other pigments.

The tone of the article is awful. The opening sentence, "Ask any artist..." is flatly untrue, this stuff is basic for most visual artists (I first learned about the CMY primaries in a community ed. Painting I class). It's not that the article is factually wrong, it's just basic stuff presented as if it were obscure or clever.

None of this makes the color wheel or any other color theory any less valid. The "four color" wheel he lists at the end is not wrong, it's just silly: you could pick any points on the wheel and their opposites and have the same thing.

sfphotoarts3• 15 years ago

You said it with more finesse, but exactly. This is just an badly written article on a subject that is clearly not understood by the author trying to grab eyeballs through sensationalism.

Using printer inks as an example was the first big cue that this was link bait to a self-aggrandizing blog.

bobds2• 15 years ago

> just basic stuff presented as if it were obscure or clever.

I think the part about the three filters is at least somewhat obscure and probably the most useful thing I learned from the article.

speleding• 15 years ago

Can you explain why your statement "pretty simple... RGB are the primaries in light" is true, his explanation in his article "complex... RGB is just a rough approximation" is false? He gives a plausible physiological explanation.

sp332• 15 years ago

RGB is pretty rough, you can clearly see from the graphs in this article, https://secure.wikimedia.org/wikipedia/en/wiki/CIE_1931_colo... how "fuzzy" all those colors are.

jacobolus7• 15 years ago

> something that is pretty simple in reality

It’s not that simple: explaining it properly takes dozens if not hundreds of pages.

> RGB are the primaries in light.

This depends on what you mean by “primaries”. As far as I can tell, your definition is based on common color reproduction technologies, rather than the physiology of human visual perception. That’s fine, but recognize then that the “primaries” chosen for practical use are constrained by economic factors, &c. The three best lights for additive color reproduction are indeed R (an orangish red color), G (a yellowish green), and B (a blue-violet color): this is because those are the colors which maximize the differential responses of different cone cells, as can be seen in this diagram in Hunt’s book The Reproduction of Colour: http://i.imgur.com/ZOdZc.png Of course, such narrow-spectrum sources are not economically/technically feasible, and so instead a typical computer or television uses lights like these: http://i.imgur.com/JHeGa.png By contrast, a typical subtractive system uses primary dyes like these: http://i.imgur.com/qri1f.png (the colored lines on the charts are the reflectances at various concentrations)

You cannot reproduce any color the human visual system can perceive through just three primary lights: every display system has a “gamut”, and for example computer displays have great difficulty displaying saturated blue-green colors.

As Hunt summarizes,

“It will be realized that these three expedients cannot correct for the fundamental limitations of the process, which spring from the nature of the colour mechanism of the eye and the shape of the spectral absorption curves of the best available cyan, magenta, and yellow dyes. What is claimed for modern subtractive processes is that they produce pleasing colour pictures, and that the inevitable inaccuracies are balanced in such a way as to be least noticeable.”

> The "four color" wheel he lists at the end is not wrong, it's just silly: you could pick any points on the wheel and their opposites and have the same thing.

This is not true. Color opponency and the specialness of the so-called “unique hues” have had a great deal of scientific literature about them (physiology, psychophysics, linguistics, etc.), and no, you could not just pick any four arbitrary points. Of course, there’s some learned/cultural component to people’s color categorization too, and there are individual physiological differences, so there’s inter-observer disagreement on precisely what color is “unique red”, etc. But it is indeed true that any color can be described as some combination of red, yellow, blue, green, white, and black: this is Hering’s theory of color vision, the inspiration for the Swedish NCS system, based on decades of rigorous measurements in the 40s–70s.

The current scientific consensus is that color vision can be modeled in a simple way with 2 stages: (1) trichromacy of 3 cone responses, (2) higher-level opponent mechanism. There are many more complicated effects beyond that, and it’s essential to consider adaptation, but those 2 cover the basics.

* * *

All of that said, I’m not completely satisfied with the original essay either. It’s pretty fluffy and hand-wavey, and the jokey language gets in the way. Conflating long/medium/short cones with red/green/blue colors is dangerous because it hides what’s really going on. I wouldn’t, as the author of this article does, call the red–yellow–blue–green anchored hue circle “proper”; there are other equally valid organizations, such as the Munsell system’s, which aims for perceptually uniform hue spacing. Still, on the whole it’s on the right track.

* * *

I’m sorry that the Wikipedia articles about these topics aren’t clearer and more comprehensive, or I’d point you there. As the article says, the best resource online is Bruce MacEvoy’s handprint.com, but several books have excellent explanations. If you’re interested and MacEvoy’s site doesn’t clear things up I can suggest where to look in the library.

cynicalkane• 15 years ago

This isn't correct. The color space is horseshoe-shaped and cannot be represented by a triangle of three additive primary colors. In addition, standard color gamuts tend to be quite a bit smaller than theoretically possible. Check out: https://secure.wikimedia.org/wikipedia/en/wiki/Gamut . This is why hi-def TV is defining extensions to the usual colorspace.

Furthermore, the laws of physics don't particularly care that humans perceive three color dimensions, and surfaces and lights are allowed to absorb or emit whatever wavelengths they please. When you have a light source with a funky spectrum, this can cause colors to look very strange, as a surface might not reflect a funny, spiky spectrum the same way (w.r.t. the human 3-dimensional color space) it will reflect a smooth color spectrum of the same color. You've probably seen cars seemingly change color when the light source is primarily those orange-ish streetlights. My parents' red minivan, for instance, turns a red-tinted grey.

slavak• 15 years ago

Sorry to disappoint you, but color theory is in fact anything BUT simple. For starters, RGB are not "THE primaries in light." In fact, primary colors aren't a property of light but of the way you perceive light. You could just as successfully choose a different set of 3 primary colors and still be able to represent [an approximation of] any other color. That is, of course, assuming you're trying to represent color to a human with 3 types of cone cells - otherwise things only get more complicated...

podperson2• 15 years ago

I think you're right in some respects (e.g. most artists have a much more sophisticated understanding of color than red, yellow, blue -- that's the version you get taught in elementary school).

But you're missing the point in others: half the article is devoted to explaining the inadequacies of the RGB model for handling actual real world color. (This is why photos of sunsets -- digital or film -- never look right.)

A simple example -- color looks weird under "white" LEDs (at least the current ones) because they're actually RGB LEDs balanced to create the illusion of "white" light. Some orange things will look all but black under a white LED while others will look orange. Why? Because orange light can be actually orange, or a mixture of wavelengths that gets a similar response from your eye.

And so on and on. The article is a bit annoying (and it would help if it assumed most readers will know about CMYK and RGB color models already) but the fundamental lesson -- that color is more complex than you think and you need to understand the underlying physics and physiology to really understand color is worth making.

jacobolus7• 15 years ago

> This is why photos of sunsets -- digital or film -- never look right

The main reason sunsets don’t look right is that sunsets have a huge dynamic range, beyond the ability of our printed photographs or computer displays to reproduce (no one has the sun in their living room).

olavk• 15 years ago

The content of the article is a fine introduction for the layman, which is framed with the stupid and condescending "Artists gets it wrong" angle.

It would be like an introduction to the web which begins "professional software developers believe that the internet is the same as the blue 'e' icon on the desktop, but actually..." No, software developers does not believe that, but the layman reader may.

Btw. I disagree that mixing primary inks yield brown because inks are imperfect. Rather, brown is the color you would expect according to theory, because mixing pigments should average their effect.

kylec• 15 years ago

I must have missed something as a seven-year-old, because this is the first I've heard that blue and yellow were supposed to be opposites. To me, it always made sense that blue and orange were opposites, both conceptually and visually. The same with purple and yellow.

Zaak• 15 years ago

It's possible you're an anomalous trichromat: http://en.wikipedia.org/wiki/Color_blindness#Anomalous_trich...

KevinMS2• 15 years ago

Yes I don't understand why this mistake is being made. The opposites are blue-orange, purple-yellow(easter colors), red-green(christmas colors).

jessriedel2• 15 years ago

EDIT: joeld42 answered my question elsewhere on this page:

http://news.ycombinator.com/item?id=2166782

The artists reds and blues are just approximations to the magenta and cyan of printers.

----

So I understand that in order to mimic the electromagnetic stimulation to the cones you only need to control the amount of RGB light hitting the retina. And I understand that propagating light is additive, so that for light producing displays you mix amounts of RGB, whereas pigments subtract light, so that for printing you mix amounts of CMY. (C paint absorbs non-R light, M paint absorbs non-G light, and Y paint absorbs non-B light.)

I further (kind of) understand that once printing and CRT screens are explained, there are additional properties of light due to the filters described in the article.

But can someone explain why the artist's paints are different from the CMY ink pigments? It seems like it's either (a) something weird about the way paints mix, (b) artists use paints which somehow take advantage of the filters in the eye, or (c) a combination.

lutorm• 15 years ago

Maybe you mistyped, but cyan paint absorbs red light, it reflects everything but red, i.e. cyan.

jessriedel2• 15 years ago

yes I did

kronusaturn• 15 years ago

Partly because cyan and magenta pigments good enough that you can mix them with yellow and get a decent red and green are a relatively modern invention. Partly because the reds and greens you get still aren't nearly as vivid as unmixed pigments. Partly because it's easier to start with a pigment that's close to what you want and adjust it than to start from scratch with only CMY.

bobds2• 15 years ago

Site was temporarily unavailable when I tried to post a comment, so here it goes:

This post clears up a few misconceptions for me, thanks for that.

I think you will enjoy the following link about "context" related illusions: http://www.psy.ritsumei.ac.jp/~akitaoka/color12e.html

Akiyoshi Kitaoka is a university professor from Japan that has created a very extensive and amazing collection of illusions, documenting so many subtle ways our eyes play tricks on us.

His homepage is really worth exploring:

http://www.psy.ritsumei.ac.jp/~akitaoka/index-e.html

nollidge• 15 years ago

I'm getting a 503 error. Obligatory Google cache:

http://webcache.googleusercontent.com/search?q=cache:oS8AaG3...

martian• 15 years ago

In the early 20th century Albert Munsell did a series of empirical studies to determine exactly how people see colors relative to one another. The results are expected, but fascinating.

http://en.wikipedia.org/wiki/Munsell_color_system

jacobolus7• 15 years ago

Why are the results expected?

(Also, I really need to expand the Munsell Wikipedia article, but Albert Munsell was dead before much of the empirical work that went into the 1929 Munsell Book of Color – by that point his son was running the company – or into the 1943 Munsell Renotations, which were based on the work of the Optical Society of America.)

mattmillr• 15 years ago

It's interesting that Red/Green and Blue/Yellow colorblindness correlate to the opposite color pairs that Jason points out. Does anyone know why this is? Are these types of colorblindness related to a deficiency in the filters (#s 1 and 2) he describes?

knutae• 15 years ago

I find the CIE Standard Observer graph enlightening: http://en.wikipedia.org/wiki/CIE_1931_color_space#Color_matc...

As far as I understand this, the red function, which corresponds to the red receptors in a typical human eye, reacts mostly to high wavelengths, but also has a small spike in the lower end of the spectrum. This explains why the violet end of the spectrum looks similar to red to human eyes, and this is probably why color wheels seem so natural. After all, a color wheel is almost the same as rainbow that wraps around.

I also find it fascinating that the CIE color space was defined as early as 1931.

jacobolus7• 15 years ago

No, the CIE standard observer is not the same as the responsivities of cone cells in the eye. That article you linked describes how it was arrived at:

1. Match monochromatic (narrow wavelength band) light sources of each wavelength with three particular monochromatic lights (called “R”, “G”, “B”), in some cases needing to add some of one of those three to the monochromatic source in order to get the two sides to match – this is a “negative” component. This process results in three (r(λ), g(λ), b(λ)) functions of wavelength.

2. Take three linear combinations of those three functions and call them (x(λ), y(λ), z(λ)), such that (a) the y(λ) function is approximately the same as the photopic luminance function as best it could be computed at the time, (b) all three of x, y, and z had only positive values, and (c) integrating each of the three yields the same value, so that an equal-energy stimulus will have the same X, Y, and Z values.

The red receptors in the human eye do not really have this kind of spike in response in the lower end of the spectrum.

* * *

However, “red” the “psychological primary” or “unique hue” does have some violet/blue in it. Any monochromatic red light source is a bit on the orange side.

steadicat• 15 years ago

Why does he conclude that we need 4 primary colors? He knows that our eyes have three kinds of 'sensors', which roughly correspond to [R, G, B]. How the brain processes the initial perceptions, allegedly [R-G, (R+G)-B, R+G+B], doesn't change the fact that you can approximate all colors by mixing quantities of R, G and B.

What I'd find more interesting is a proposal (or a mention) of a color space that's based on what are, according to him, the 'computed' values. Something like NTL (tiNt/Temperature/Luminance), where N is R-G, T is (R+G)-B, and L is R+G+B. (The names 'tint' and 'temperature' are taken from photo editing software, as they are the only tools I can think of that come close to this system.)

cwp• 15 years ago

Actually, this kind of encoding is quite common. Analog television, digital video and JPEG images are all encoded using brightness, red-green and blue-yellow channels. The details vary (eg. between YUV, YCbCr and YPbPr) but they all take advantage of the fact that our eyes are more sensitive to variations in brightness than to variations in colour. As a result, we can subject the colour data to higher levels of compression without noticing any visual degradation of the image.

armandososa• 15 years ago

     Artists get it wrong

I kinda get upset about all his use of "artist" as a derogatory term. Have he ever painted anything in real life? Beleive it or not, just a canvas plus red, blue and yellow acrylic paint will be enough.

Oh, and some kind of talent.

jackolas• 15 years ago

I'm not sure that understanding color as we experience it is required to be a good artist, but it might help the more engineer-ish of us.

icegreentea• 15 years ago

Agreed. The fact that artists make stuff that 'works' at all should be hint enough that they aren't completely wrong. Just working with a useful approximation/abstraction layer.

KevinMS2• 15 years ago

Beleive it or not, just a canvas plus red, blue and yellow acrylic paint will be enough.

This is actually not completely true, but I appreciate the point you are making.

Something we shouldn't overlook is that there no pigments that fit perfectly on a color wheel. In other words, there's no blue pigment, its always blueish green or blueish purple.

This has been a challenge for pigment makers for centuries and has a complicated history.

Maybe sometimes a perfect pigment or dye is found, color wise, but then it turns out to easily fade, aka 'fugitive', so artists, the ones who expect their art work to last for more than a few years, are stuck with certain colors.

Why does this matter? Its because of mixing. You might think you have paints of the primaries, blue, yellow, and red, and you need a bright green, so you mix the blue with the yellow, and you end up with an muddy olive color.

Turns out your blue was leaning towards being purplish, and when mixed with the yellow, a component of grey was produced, because they are opposite.

So how do you get a nice bright green? You need to mix a greenish blue with a greenish yellow.

In practice this results in artists, who want to paint with a full range of colors, need 6 primary colors.

For example, two good lightfast (not destroyed by light) blues are ultramarine (very old pigment), and a newer phthalo blue (much more modern). The ultramarine leans towards the violet and the phthalo, to varying degrees, towards green. In the old days I think prussian blue was the greenish blue of choice, but its a weak color and is easily overpowered.

Cobalt blue, if I'm not mistaken, is the closest thing we have to a lightfast primary color.

Artist really hit the jackpot with blues. Some are very old. They are also nicely transparent, which can make a difference in technique

The bright reds and bright yellows aren't so good, and the ones that are lightfast are supposed to be toxic. And not very transparent either.

If you don't care about the brightness of your reds and yellows you are in luck, there are plenty of rust and dirt pigments like iron oxide, sienna and umber.

This is a complicated subject. Here's a book about it

http://www.amazon.com/Blue-Yellow-Dont-Make-Green/dp/0967962...

tintin• 15 years ago

Well even black and white could be enough ;)

But seriously, I also paint with CMYK only and there are a lot of tricks to get colors you cannot create with these colors alone. Like bright red. When you first paint yellow, let it dry and paint red over it (transparent) you will get some sort of light bending which makes the red brighter. Car painters do this all the time.

But the author is right in a way. By creating contrast you can paint black darker then your paint is. It's all about what's going on in the mind.

m-photonic• 15 years ago

>And magenta? It comes from full R and B with no G, activating Filter #1 full-positive, Filter #2 at zero.

This doesn't seem right to me. If the second filter were at zero, you should have a pure red and not something with blue content in it like magenta clearly has.

I think he may be representing the second filter as R+G-B, when R+G-2B would make more sense. The latter system shows FFFFFF as being neutral on the yellow-blue axis, while the former erroneously puts it in the yellow region.

jacobolus7• 15 years ago

None of these are straight-forward “linear” sums: there are differing amounts of each type of cone cell in the retina (and the proportions vary from one part of the retina to another), there are several levels of combination of signals which we don’t fully understand currently, the eye/brain adapts to what it’s just been looking at, what else is in the visual field, what it knows the light source to be, what “memory colors” it expects for an object, and so forth.

Thinking of the mechanisms of the eye operating directly on “FFFFFF” is a very imprecise model for what’s happening.

apl2• 15 years ago

  > But I digress, and besides I did promise to be all gross
  > and irresponsible, so I’ll stick with that.
  > So there are R, G, and B cones.

Cute and all, but so very wrong. There are acceptable simplifications; this is not one of them.

mambodog• 15 years ago

A while ago I submitted an article[1] about the YUV colour system, which takes into account human perception of brightness/luminance, and how it differs between colours. It's also an interesting colour system to work with, as the 'Y' channel just stores luminance, so discarding the other two channels gives you a greyscale image.

I recommend reading it if you haven't already.

[1] http://nreynolds.co.uk/blog/hsv-is-dead/

blahblahblah• 15 years ago

So there’s no such thing as "red with a little green" -- there’s just a less intense red. The brain physically cannot see "greenish-red" because the filter removes that information.

Under ordinary circumstances this is true. However, it has been shown that the human brain can be made to see reddish-green, yellowish-blue, etc. Scientific American published an article about this strange hack of human visual perception just last year. http://www.scientificamerican.com/article.cfm?id=seeing-forb...

thret• 15 years ago

If you enjoyed that article, you may also like http://www.designersreviewofbooks.com/2010/10/interaction-of...

It is a beautiful book.

Color wheels are wrong? How color vision actually works

Related Stories