The commonest form of color wheel shows the three subtractive primaries, Red, Yellow, and Blue, with equal spacing.
The Munsell system of colors, which I think is one of the best color systems, uses a circle with ten hues: the five colors Red, Yellow, Green, Blue, and Purple are equally spaced.
And I remember a news item where Red, Yellow, Green, and Blue were identified as the "psychological primaries".
And of course color television and your computer monitor uses Red, Green and Blue as additive primaries, and color printing, unlike color painting, terms the subtractive primaries Magenta, Yellow, and Cyan.
Well, here's my little addition to the general confusion. The color circle below is based on my notion of how to subjectively space the colors generally acknowledged as primaries:
Red is closest to Yellow. Blue is further away from Red, since those two colors stand at opposite ends of the spectrum. And Blue is farthest from Yellow, since Green is percieved as almost a primary: and the distance between Blue and Green equals the distance between Green and Yellow, and both are less than the distance between Red and Yellow.
A nice color circle meeting those conditions with 60 divisions can be produced by using the sequence of numbers 4-5-6.
Red to Orange and Orange to Yellow are each 8 hues apart.
Blue to Purple and Purple to Red are each 10 hues apart.
Yellow to Green and Green to Blue are each 12 hues apart.
And here's the color circle I arrived at on that basis:
I've modified the scheme considerably since starting these pages; here is the revised color circle, with 240 hues in the color circle:
it acknowledges that the green phosphor in a CRT should be considered as representing quite a yellowish green, for example, compared to my initial attempt.
Now, before continuing, I should address one objection that some people have to color circles like the one above.
Pure light with a wavelength of 650 nm is red; pure light with a wavelength of 575 nm is yellow; pure light with a wavelength of 475 nm is blue.
There is no wavelength of light which, by itself, looks purple. Yet purple is on the fundamental circle of saturated hues used by artists. Are they wrong: is purple only a "mixed" color, like brown or beige?
To answer this question, we need to know something about how human color vision works. The retina has two kinds of receptors, rods and cones. The rods are primarily sensitive to blue and green light, and are used for night vision. For normal vision, the cones serve, and there are three kinds of cones. Each kind is sensitive to a range of colors, but each one has a different peak of sensitivity.
We determine the color of objects we look at, or light we see, by means of the ratio between the stimuli experienced by the three kinds of cones. That isn't the whole story, because the brain performs sophisticated processing to correct for variations in the overall color of a scene due to changes in ambient light, for example, from noon to sunset, or from outdoors to indoors, under either incandescent or fluorescent light, so that we can consistently recognize objects by their own colors. But it is the starting point for color vision.
If one uses a triangular chart like this:
to plot the relative intensities of the stimuli to the three types of receptors for the different frequencies of light, one will get a curve going around the point representing the proportions of stimuli caused by white light. This is not commonly done for the actual sensitivities of the three pigments in the cones of the eye, but I finally found an example of that here, but long before those sensitivities were known, studies of the human visual system involving the matching of colors formed in different ways allowed it to be described in terms of three postulated receptor types. The first of the two major studies leading to the first of the modern colorimetric standards was conducted by W. David Wright and involved 10 subjects whose color vision was measured, and the results were published in a 1928 paper; an additional study, with 7 subjects, conducted by John Guild, had its results published in a 1931 paper. If the results of those studies were accurate, conversion between the postulated receptors and the actual receptors would be a simple linear transformation.
If one considers that hue refers to the direction from white in which a color lies, one will note that since:
then, new hues that cannot be produced by any combination of a pure spectral light with white light can be produced by combining red light and blue light.
The CIE chromaticity diagram, although it is based on tristimulus values obtained by experiments involving subjects comparing colors, and not directly on the spectral sensitivities of the three pigments found in the cones of the retina (which are now known) illustrates this principle, as it consists of an arching curve around white, with a straight line between red and blue accounting for the hues normally thought of as purplish.
Thus, if one wishes to be able to describe colors, on the basis of three co-ordinates; one being brightness (luminance), one being how colorful or gray a color is (saturation), and a third being which saturated color a color tends towards (hue), one cannot describe all colors unless one's circle of hues is complete, and if it only includes spectral colors and is missing purple, then one has cut a wedge-shaped chunk out of the CIE chromaticity diagram - which is considered to be scientific.
So purple really is one of the fully saturated colors. Don't let anyone tell you otherwise.
I am not, here, in any way denying the scientific fact that purple light must be a mixture of light of different wavelengths; I am merely stating that when red and blue light are mixed, the result can be something novel in its attribute of hue, while mixing light of one wavelength with white light only produces less saturated colors having the same hue as the monochromatic light used to start from.
Purple has even been an important color in history. I remember one day hearing a news report that Princess Diana had committed a serious faux pas by attending a function at which Her Majesty the Queen would also be present, by wearing a purple dress thereto. My initial reaction to the report was one of bemusement, but then I remembered how, at one time, the only purple dye available was Tyrian purple, made from a rare mollusk, and the use of clothing dyed with this substance was reserved for the Roman emperors.
It was not until 1856 that the dye mauve was synthesized, and the new freedom in the use of color this provided led to the 1890s being known as the "mauve decade". People have always been fascinated by pretty bright colors, and in addition, they eagerly seize on the new freedoms offered by technological advances, not only using it, but all too often overusing and abusing it; those who remember bad synthesizer music, or laser-printed documents with ten typefaces on a page, will understand: thus, we should not be too surprised that the discovery of the aniline dyes could turn a whole decade purple.
A larger version of the CIE chromaticity diagram, showing how I created the representation of it seen above:
The diagram above is a crude rendering on my part of the normal 1931 CIE chromaticity chart, but instead of being plotted conventionally in Cartesian coordinates of x and y, x, y, and z have been treated as three items that must sum to 1, and are shown in an equilateral triangle. Note that an elongated triangle in the center of the diagram represents what is called a Maxwell triangle, showing colors derived from mixtures of light from the three phosphors of a typical monitor; outside of this triangular area, the gamut of a monitor, the colors on the periphery are extended to the perimiter on the diagram following curved lines of perceived hue shift with saturation change.
As this diagram was developed by the Commission International de l'Éclairage to characterize lighting, the position of a source of light on this diagram depends solely on its relative spectral composition, and not on its brightness. This means that distance from white at the center represents spectral purity, not saturation or chroma. A consequence of this is that the color of an object which only reflected light with wavelengths from 475 to 476 nanometers, which would look nearly black in normal light, would be assigned a position near the edge of the CIE chromaticity chart, corresponding to the vivid color it could have if subjected to an extremely bright source of incident light. This also means that it is possible to plot curves of the maximum achievable spectral purity for the color of objects seen by reflected light based on the proportion of incident light that they reflect; these curves are called the MacAdam limits.
Relating the mixtures of red, green, and blue phosphors on a CRT to the colors of pure spectral light is difficult, but based on information I have found, this chart may have some approximate validity:
In addition to a color circle of fully saturated hues, there are also, of course, less saturated colors.
The following diagram includes a full gamut of such colors, with only 30 hues in use. To make it easier to calculate the RGB values for the palette I needed, and to make it easier to draw the diagram in which the colors were placed, I did not attempt to make vertical planes of constant brightness, but instead simply moved from the circle of fully saturated hues towards white and black, somewhat as is done in the Ostwald color system (but I use squares instead of hexagons, having 1, 3, 5, 7... colors instead of 1, 2, 3, 4... in successive shells):
For a gamma of 0.85, this illustration shows a color chart made up of colors in that arrangement: the diagrams, in three rows, show every second hue, starting with red, for 24 different values of brightness, excluding black and white, and 9 different values of saturation, excluding no saturation. The diagrams have ten columns, however, because the gray scale is included with each diagram.
This diagram is in the PNG file format, which is not supported in some older browsers. It has been necessary to use that format for this file, as JPEG did not give acceptable results without an excessive file size, and it has more than 256 colors.
Note also that this color scheme is not the Munsell system, despite a superficial resemblance. In addition to the different arrangement of the hues, no attempt is made to correct for apparent hue shift for less saturated colors.
This page, if viewed in 256 color mode, will simply look less attractive, due to dithering. As background colors are not dithered, the page with the more extensive gamut will simply show incorrect colors if viewed in that mode.
Because on many computer systems, RGB values are not linear specifications of brightness, it is necessary to take into account gamma correction to present colors accurately. The charts on this page, except for the chart of "internet-safe" colors, which need not be adjusted, as browsers display those colors without correction, have been corrected by being raised to the exponent 0.85.
The following table should allow you to determine the gamma correction required for your computer display. The gray squares attempt to achieve the same shade of gray directly, and by alternating white and black pixels (or full-intensity red, green, and blue pixels in one case): where the vertical bars in each square are the closest match in apparent brightness, the correct gamma for your monitor is found.
The first item in each row is the chart for checking gamma. The second item is my color circle, presented correctly for systems requiring the gamma correction indicated by that row. The third item is the gamma value (actually the reciprocal of the gamma of your monitor):
 |
 |
1.00 |
 |
 |
0.90 |
 |
 |
0.85 |
 |
 |
0.80 |
 |
 |
0.70 |
 |
 |
0.60 |
 |
 |
0.50 |
Of course, there are other ways to organize a gamut of colors that includes possible colors of all kinds to at least approximate any color. The "internet-safe" colors that work in 256 color mode on common browsers are an example of what is known as process color, and are shown here:
This can be thought of either as a chart of additive colors, or as a chart of subtractive colors. Starting from black, one goes up to add blue, to the left to add red, and away from the observer to add green. Starting from white, one goes down to subtract out blue light using yellow ink, one goes right to subtract out red light using cyan ink, and one goes towards the observer to subtract out green light using magenta ink.
However, in fact, the chart depicts colors produced by addition from monitor phosphors, and not subtraction. Also, in practice, while yellow ink indeed looks yellow, cyan is more like a light blue, and magenta more like red than like purple, than the colors in this diagram.
A similar process color chart with colors more closely approximating what might be available through printing is shown below:
By default, Microsoft Windows uses a different palette of process colors for bitmap images in 256 color mode. Instead of using only 216 colors, all 256 colors are used: the brighter components red and green vary through eight steps rather than six, while only four levels are allocated to blue. This produces a range of colors having the following appearance:
The colors chosen tend to favor the darker colors: for blue, the levels are 0, 64, 128, and 255, with 192 omitted, and for red and green, the levels are 0, 32, 64, 96, 128, 160, 192, and 255, with 224 omitted.
Favoring the lighter colors somewhat instead, and allocating four levels to blue, five levels to red, and twelve levels to green, using:
0 28 57 85 102 119 136 153 170 187 204 221 238 255
-----------------------------------------------------
Blue: 0 119 187 255
Red: 0 102 153 204 255
Green: 0 57 85 102 119 136 153 170 187 204 221 255
one obtains the following chart:
In addition to the 240 colors formed systematically, the missing gray shades with 28, 57, 85, 102, 119, 136, 153, 170, 187, 204, 221, and 238 for all colors are added. This leaves only four colors unused.
In the main section of the diagram, the squares whose borders are black instead of gray are those which would be placed on the circle of maximally-saturated colors available within this gamut.
The three images below:
show the results when a color photograph is rendered according to this color gamut, first by mixing colors in adjacent points to approximate actual colors in the scene not present in that gamut, and then by strictly replacing the color at each point by its nearest available neighbor.
The nearest-neighbor image is clearly not perfect; however, for comparison, here is a nearest-neighbor image using the 240 "internet-safe" colors,
and one using the 256 color palette from Microsoft:
Of course, the choice of nearest neighbors by the paint program used may also affect the results, and my scheme has a particular advantage resulting from adding a gray scale to the gamut. If the gray scale were omitted, a nearest-neighbor image using my scheme would look like this:
which is not nearly as much of a visible improvement over the results produced by the other two palettes as the earlier image, although it does seem to me to still be some slight improvement.
For comparison, here is a nearest-color image using the color gamut above that used a stack of color circles based on a 30-hue version of my color circle:
Just as the circular chart that began this section represents the color world as seen by a painter, who has a palette of oil paints representing several saturated colors and also white and black paint, the process color charts represent the color world experienced by someone working with a color monitor, or with printing.
To make a particular color, a painter should use the two saturated colors nearest in hue to the desired color, plus as much white or black paint as is needed to make a good match.
When printing a color photograph by conventional color separation techniques, not involving the use of computer-generated halftones, it is necessary to use cyan, magenta, and yellow to produce the various colors. A light black-and-white image of the same photograph is also overprinted.
But ideally, if one is trying to produce a desired color directly, one should mix at most two of cyan, magenta, and yellow inks with as much black ink as required. For almost the same reason that a painter should not try to create a desired color using widely separated hues. In the painter's case, too much trial and error is needed that way; in the printer's case, the limitations of the inks used are more important.
Color Atlas by Harald Küppers featured a set of color squares for this kind of system; it was an inexpensive small book with a silver cover. Below is a diagram of what a process color chart looks like in that kind of system.
The first row shows colors built from yellow and magenta; the succeeding squares going to the right have more and more black added; first 25%, then 50%, and finally 75%. Since having 75% black added to the colors makes them difficult to distinguish, instead of a complete square with 25 different colors, only nine separated squares, each containing one color, are shown at the end of each row.
The second row shows yellow and cyan, and the third row magenta and cyan.