h2g2 The Hitchhiker's Guide to the Galaxy: Earth Edition

Newton may have discovered that the rainbow consisted of red, orange, yellow, green, blue and violet but, when it comes to colours (rather than light colours), orange, green and violet are NOT primary. They are secondary colours because they are formed by mixing two other colours together - red and yellow, yellow and blue, and blue and red respectively. The only tertiary (?) colour, found by mixing all three primary colours, is brown.

I hate to be pedantic, but ... (Actually, I love being pedantic!)

Quite right. That's one place where I used the word 'primary' in a colloquial rather than technical sense. Should I have called the colours of the rainbow 'basic' colours?

I did have to make some conscious decisions about what to exclude; it would have been very easy to start talking about additive colour, too. But I made the deliberate decision to try and limit this ramble to colour on computers.

Lil

This is all very interesting to a color-challenged person like myself (married to an art major). In fact, I greatly prefer black and white photography, its not as confusing - all the Photoshop tools for "color correction" are mystifying, what is "correct"?

One approach that makes some sense to me is a little book "Color Harmony 2 - Featuring the Palette Picker" by Bride M. Whelan. The blurb says "a guide to creative color combinations". It starts out with the primary (3), secondary(3) and tertiary colors(6), then a "color wheel" that takes each of the 12 color "segments" and adds 8 tints/shades, for a total of 96 colors. Then the theory of combinations is laid out - analogous, clash, complementary, monochrome, etc. - and how to pick the colors off the wheel starting with any given color. Another more subjective division is given - magical, energetic, professional, elegant, calm, etc. So sort of a cookbook. And lots of real world examples like fast food restaurants (orange projects an inviting message of good food at inviting prices - Burger King!). The included CD allows rapid generation of "feasible" combinations.

I'm still trying to understand why anyone needs the 64 crayola box, 16 is more than I can handle

Ah, but orange green and violet don't *have* to be made by mixing other colours. It is possible to emit light which is pure orange. (I mean 'pure' in the sense that it is at only one wavelength.)

The idea of primary colours comes from an artifact of the way our eyes work: it is possible to generate all colours the eye can see by mixing together just 3 different colours in varying proportions. This is because our eyes are only actually sensitive to 3 different colours. It so happens that those colours are red, green and blue, and for illumination, you can form what appears to be the whole gamut of colours by mixing these. Inks work by effectively filtering out light, so we have to use the pale counterpart 'primary' colours here: cyan (pale blue), yellow and magenta (sort of pink).

Whilst it is true that you can make 'orange' by mixing red and yellow (which means mixing 2 parts of red light to 1 part of green light, or alternatively 1 part of magenta ink to 2 parts of yellow ink), this is actually very different from a 'pure' orange. It's just that our eyes will be unable to tell the difference. The limitations of our eyes however doesn't make 'orange' any less of a colour than blue.

I rest my case. The distinction between our intuitive ways for describing colours and the technical terms for same are almost as bad as the terminological snarls of philosophy.

Steve, there is nothing wrong/inferior with your book on colour harmony, but it doesn't completely bear on the topic of this entry, which is DIGITAL colour. What you describe has to do with pigment-based colour, which is what painters use. Pigments are suspended in water or turps-soluble media to create watercolours, oil colours and acrylics, or pigments are ground up with chalk to make pastels and so forth. At the creation stage pigments are added to each other to make intermediate shades ... or mixed on the palette to the same end ... or scumbled on the drawing to the same end.

But pigment-based colour systems follow different rules than light-based models.

IanG, no colour is inferior to any other colour, and I hope it would be understood that if I'm talking about web-safe colours then I won't be referencing any out-of-frame wavelengths, pure as they may be. But to be sure, I'm not sure what your terms are here. Your argument seems to mix several quite different colour models. Could you give an instance of something that emits a "pure orange" wavelength? Outside of spectrum analysis, which I know little about, I thought orange (as the ICC seems to believe) is either a red-green mixture in light modelling, a red/yellow mixture in additive models, and yellow-magenta in printing. What orange are you thinking of?

"But pigment-based colour systems follow different rules than light-based models."

Curiouser and curiouser. I guess I should admit I vaguely realized that fact when I posted my (partially tongue-in-cheek) message. The "Web Safe Colors" writeup mentioned that Pantone is a member of the ICC (Adobe, MS, Apple, etc.), even tho that co. is more on the pigment side than the digital, in an effort to maintain some connection.

I should also admit to topic "extension" if not downright drift, as I am more concerned with what color combinations "work", regardless of how they are created (digital, pigment, ...) For example, my little books's "Clash Scheme: Combines a color with the hue to the right or left of its complement on the color wheel" would seem to hold in either digital or pigment terms ... or not. But I drift further, sorry, back to my grayscale images

"Scumbled", a new word for me

I was indeed talking about several different colour models - I guess the thing about colour is that there are lots of ways of looking at it, and some make more sense than others. (And I didn't actually mean to disagree with anything you said, just to point out that there are other points of view.) Also, all of the colour models I'm aware of (RGB, CMY, CMYK, HSV and all its variants (most noteably HSL) and the one whose name I can never remember but which defines white in terms of the radiation given off by heating something to a specific temperature) are equivalent in the sense that you can convert between them. (Although some have a wider gamut, and can hence express a wider range of colours than others.)

An example of nearly pure orange is given in Joanna's fine article here: http://www.h2g2.com/A337259 on absorbtion and emission lines. Of course some people would argue that sodium vapour lamps are yellow, not orange, in which case I'll go with orange LEDs instead. (They're unusual but you can get them.) I say 'nearly pure' because it's still a mixture of two colours, but they happen to be almost exactly the same shade - it's *not* a mixture of red and green, it's a mixture of two very similar shades of orange! (This is why some colours look very strange under these street lamps - I had a red car which gave a good impression of being green under these lamps!)

Actually the best example of pure orange is the orange bit in a rainbow, or in the spectrum that comes out of a prism.

This is all very well for physicists, but not a lot of use to anyone trying to produce colours on a computer screen or by mixing inks or paints.

Physics! *restrains urge to run in opposite direction* I know of Joanna's articles as I had the honour of subbing one; they are nearly as good as her doughnuts. I'll see whether she wants to add comment here.

I would have liked to talk about gamut, too, because that's another one of the weird and interesting asides in colour theory, that some models yield more colours than others.

Thanks for expanding the discussion.

The thing is, orange emission lines in sodium are all very well but you're not going to build a monitor which works by having some of every element in gaseous form in each pixel, to be excited to give off a colour...

You have to compromise- I guess at that point it moves from the realms of physics to the biology of what we actually see, and the realms of RGB...

Occasionally, mention is made of colour blindness - and this is actually fairly important to this topic. Colour blindness comes in two flavours - one is extremely rare and can happen to both males and females and one has only monochrome vision. The other is often known as a colour deficiency and in its most extreme form the colour blind person has only two types of cone on the retina - those for sensing red and those for sensing green or blue as opposed to normal colour vision when a person has a mixture of all three sensing cones. This form of colour "blindness" only occurs in men (it is a sexually-linked genetic trait) and at a rate of about one in 10 men carry this defect. We (yes I am red-green colour blind ie. missing (I think) the green cones) can still see a full spectrum of colours but are incapable of fine discrimination of a few shades. With red-green, the problems come around olive-green, brown colours where whether I would name something as brown or green becomes almost completely arbitrary and leads to some interesting colour combinations in the clothes I choose to wear (although I tend to avoid buying anything with these colours these days).

However because of the eye's ability to cope with only two inputs for colour it would be in principle possible to make colour displays by mixing instead of three colours simply two colours, and the eye would be satisfied with the result - at least for 10% of the male population!

So the key point here is that 'true colour' really just means 'good enough for human vision'. The fact that we can get away with mixing just 3 colours together is purely because our eyes only discriminate 3 colours, rather than being capable of a full spectrum analysis as they clearly should have been.

Actually I had an argument with my lecturer in computer graphics at college - he pointed out that the colour receptors our eyes don't respond to a single wavelength - if you plot their response against wavelength, you get sort of hump, centered around the colour. The three humps, for red, green and blue all overlap. His theory was that you shouldn't therefore be able to accurately recreate all colours with just 3 input colours in the first place.

I didn't buy this because the output from the eye when presented with a single 'colour' is just 3 values, and so you shouldn't need more than 3 inputs... However I think in retrospect he may have had a point as a result of the overlapping colours - I think my solution may have involved picking a 'red' and a 'blue' that were almost out on the infrared and the ultraviolet, but just within the eye's range of sensitivity so that I could adjust these without interfering with the green levels. In real life, this is not how we build colour displays.

The point about overlapping colours would explain the failure of the human eye to distinguish the metameric greys.

As a footnote to colour-blindness, I understand that the 3 lights in traffic lights have been selected for colors with different contrast values so that even the monochromatically colour blind can learn to distinguish between them.

IanG, Is what you were thinking of the colour temerature (something usually measured in Kelvin)?

*is scared*

OK so if we assume that the colour receptors for each of red, green and blue only are active for a non overlapping range of light wavelengths (colours). The determining factor how many colours we can see is then limited by the number of intensity levels at each of the three colours.
This assumption (like all assumptions) is very flawed, as has alredy been said, the receptors are not brick wall filters, so they overlap. The eye is also not a perfect piece of optical engineering so it distorts the colours slightly. The receptors probably have different intensity responses (to the extent that one or more may not work at all leading to colour blindness).

*Trying quite hard to think of something to tie all this together, but I can't *

*delves into memory of visual psychology course*

Yes, IanG is completely correct that it's the different types of cones in the eye which allow colour vision, but I'd like to clarify and/or confuse the issue over colour blindness.

Of the three types of colour receptor in the eye, two (red and green) are very lose variants of the same structure, the blue cone is rather different. In red/green colour blindness, it is not that one type of cone is "missing", but that the eye produces only one type of cone (somewhere between the current red and green) which may well be the evolutionary precursor of the two types. It is the gene for producing red and green cones which is sex-linked: not completely unknown in females, but they have to have two "broken" genes instead of one: if the incidence is 1 in 10 in men, it would be 1 in 100 in women.

The interesting thing is that women with one "broken" gene and one "normal" actually have *four* different kinds of cone (blue, red, green and reddish-green). These women tend to have slightly odd colour sensitivity (somewhat enhanced, but somewhat skewed), and see colours as non-matches which most people think are the same. To produce monitors which could produce any colour they can see would require four emission colours. It's easy to tell if you are one of these: it is necessary and sufficient that your father was colour-blind, and you're not.

A propos of nothing at all, pigeons also have four types of colour receptor!

It's more involved than colour temperature, Phil. There was some standardised attempt to unify various different colour models, and in particular deal with the fact that gamuts of real monitors, LCD panels and printing processes are not only different, but that none is a complete superset of any other.

So they took a step back and came up with a fairly complex formulation for colour. Colour temperature measured in Kelvin would have come into it, but that only deals with white. I can't remember how they dealt with colours.

(This is all dredged from dingy memories of stuff I only glossed over about 7 years ago, so if someone actually knows about what I'm trying to talk about please step in!)

The more we talk about colour perception, the stranger it gets. I'll never be able to prove whether I have skewed colour reception because my dad, who was adopted, has been dead for 45 years. But I do seem to get into frequent disagreements about what colour blue-green things are. I always thought it was because colour was inherently subjective. How delightfully strange to think I might have nonstandard eyeballs.