07 September 2006

More on color

In my last post about color, I discussed the inadequacies of the standard color wheel and explained why we're going to have to replace it with two things: (1) a more accurate way to describe color; and (2) a system for approximating real-world paint mixing. In this post, I'll talk about describing color. As I do so, I'll refer you to certain sections of the Handprint web site (from which I have stolen shamelessly) in case you want more detail.

Although we often still see the standard three-primary color wheel in books about painting and color mixing, it really went out of date in the late 19th century, when guys like Ogden Rood demonstrated that it pretty much stinks for describing color accurately. There is nothing in the way humans perceive light to support the idea of three unmixable primary colors (red, yellow, blue), each of which is complimentary to a specific mixable secondary color (red and green, yellow and violet, blue and orange). In fact, it makes sense to me that there are no special primary colors at all, whether the traditional artist's primaries (red, yellow, blue), the printer's primaries (cyan, magenta, yellow), or anything else.

A number of more accurate ways of describing color have been developed. Many of them are designed primarily to support the needs of the print industry, the dye industry, manufacturers of video equipment, and other commercial ventures. They are needlessly complex for our purposes. The best system that is comprehensive enough, but not too complex to be easily understood, is the Munsell color system. It was first developed in the early part of the 20th century and has been updated a few times since then, although the original structure remains. Any such system represents a series of compromises, so there are ways in which Munsell is imperfect, but overall it suits our purposes better than any other that I am aware of.

Rather than a color wheel, Munsell is built around a three-dimensional color space. This space takes the shape of an irregular cylinder. Munsell uses three properties of color: hue, chroma, and value. I described those properties in detail in a previous post.

Running up the center axis of the cylinder is the property of value. At the bottom of the cylinder is value 0 (pure black); at the top is value 10 (pure white). So, for example, a value 6.5 gray is fairly light, while a value 1.0 gray is almost black.

Running around the outside of the cylinder is a hue circle. It is defined by five principal colors (there are no primaries in Munsell). These colors are red (R), yellow (Y), green (G), blue (B), and purple (P). These are generally represented in clockwise order, starting with yellow at the top. The five principle hues have five intermediate hues in between them: yellow red (YR), green yellow (GY), blue green (BG), purple blue (PB), and red purple (RP). Within each of the hues are ten subdivisions, with 5 at the center. So 5BG is a pure blue green, while 2BG is more blue and 9.3BG is more green.

Each of the principal hues has a visual compliment that is the intermediate hue directly across from it on the circle. So the compliment of red is blue green, the compliment of yellow is purple blue, the compliment of green is red purple, the compliment of blue is yellow red, and the compliment of purple is green yellow. These compliments correspond (approximately) to how humans see color. Munsell compliments are reasonably close to actual data on afterimages. If you stare at a spot of green for a long time, and then look at a neutral gray surface, most people who are not color blind report that they see an afterimage within Munsell's red purple range. The same goes for each of the other complimentary pairs on the hue circle. You see these afterimages because of the way that cone cells on the retina work. I'm not going to describe the physiology, but it helps to know that these are real phenomena (which I am oversimplifying drastically here), not arbitrary or aesthetic conventions.

If the value parameter is a line down the center of the cylinder, then the chroma parameter radiates outward from that center line to the edges of the cylinder. Zero chroma (gray/black/white) is at the center. Moving outward are increasingly chromatic (intense) colors. So a value 6 yellow at chroma 1.5 is basically a warm grey (not very chromatic), while a value 6 yellow at chroma 15 is very intense.

There is no arbitrary maximum chroma, so the chroma scale for each hue runs from 0 to however intense that hue can get. As new, brighter pigments are developed, they are simply placed at higher chroma levels than those of older pigments. Any pigment can therefore be placed upon the Munsell color tree. Because of the physics of light and the nature of color vision, the maximum possible chroma is different for different hues. For example, the maximum possible chroma of a light-valued yellow is much higher than that of a light-valued purple. Maximum chroma for a given hue is also different depending on value. So the Munsell color space is a bumpy, uneven cylinder (when Munsell first invented this system, his realization that the color space couldn't be symetrical was a big improvement over previous systems that had tried to cram a messy reality into an idealized circle or triangle).

Colors are named in Munsell in the standard notation of hue value/chroma. So vermillion is noted as 8.5R 5.5/12. That means that, within the hue of red, it is at position 8.5 (closer to yellow red than a pure red), with a value of 5.5 (right in the middle) and a chroma of 12 (fairly intense). Some paint manufacturers, such as Liquitex, put these numbers on every tube of paint. Unfortunately, that's rare.

The Munsell system has been updated several times to make it more technically accurate, but none of those updates is significant for our purposes. You can buy color sets from the Munsell company. They consist of a book describing the system, a bunch of color chips (sets have either glossy chips or matte chips), and pages with little pockets that the color chips fit into. The idea is that you learn the system by fitting each chip into its appropriate pocket. I haven't bought a color study set, not because I'm uninterested but because they cost hundreds of dollars. Getting a set and placing all of the chips would not be a waste of time for a serious student of painting.

That's Munsell. Boy, that was a lot of explanation, even though I picked the simplest useful color system that I know of and avoided extraneous detail. Color is really complicated.

So how is Munsell more useful to a painter than the old three-primary color wheel? First, it dispenses with the confusing idea of primaries and secondaries while more accurately identifying useful complimentary color relationships. These relationships are of great value in choosing harmonious color relationships. Second, as you become more familiar with Munsell, you can begin to think about colors in terms of how they relate to each other within the color space. If you are looking at a blue wall, for example, and you are thinking in Munsell terms, you can figure out where the color lies and how to accurately describe it. What is its hue? How chromatic is it? What value is it? How do those parameters compare to other colors you are trying to work with? How do the hue, chroma, and value of the wall relate to the hue, chroma, and value of the blob of paint you are trying to use to represent it? Some artists pre-mix a set of colors on their palette in Munsell value steps. Several companies sell paints that are graded according to Munsell; Studio products sells a set of neutral grays and another set of greens, all the same hue and chroma, of different Munsell values. These are particularly useful for underpainting.

There isn't much in Munsell that helps you figure out what color you will get if you mix two paints together—that's not what it's for. What it does do is help you decide what color you are trying to get to. And for that, its really excellent.

Lots more on color in future posts.

Tags: ,


painterdog said...

this is very interesting but way to complicated.

A gray value system with going from white to black should be simpley based on color values.

That is if I have a range from cad yellow light, cad yellow med, cad orange(middle value for flesh tone, or lowest tone before shadow value),yellow ochre, cad red lt, cad red med, cad red dark, raw umber, burnt unber, cobalt blue, ultra-marine blue, aliziran crimzian, black.

If you mix gary values to these colors you will have a very good control and an easy way of mixing hues from all the colors, giving you a way of understanding value and color relationships.

David said...


Thanks for the comment. I'd like to emphasize that Munsell is a system for describing color. It isn't a good model for mixing pigments. I'll get into color mixing, including (as you suggest) strategies based on pigment values and adjusting chroma with grays of equal value, as I continue this series of posts.