Hal Glicksman
White is Green

Background

I reread books of art instruction looking for the precise place in the text that physics and art diverge. This is a typical example from Craig Denton's1 otherwise excellent essay on color theory:

Color circles or color triangles are graphic vehicles used to organize colors. We will be using the artist's color circle, which relates to pigments because graphic designers generally follow its traditions and rules. There is another color wheel, composed of light primaries and their subtractive color complements, which we will discuss later. But, subtractive colors are obscurely named and physically transparent, designed to be overlayed in printing, and wind up looking unpleasantly acidic by themselves. A subtractive color circle isn't aesthetically pleasing so it doesn't help you understand how to best compose colors.

Luigina De Grandis, in Theory and Use of Color, summarizes the frustration of reconciling artist and scientific color. She defines magenta as "a red tending toward purple" and then says: "...the artist, not bound to the industrial primaries, may continue to use the reds, yellows, and blues of his or her choice." 2

Johannes Itten's statement 3 is important because his work is considered authoritative by artists, and because he studied and wrote about Ostwald in an attempt to reconcile art and science:

One essential foundation of any aesthetic color theory is the color circle, because that will determine the classification of colors. The color artist must work with pigments, and therefore his color classification must be constructed in terms of the mixing of pigments. That is to say, diametrically opposed colors must be complementary, mixing to yield gray. Thus in my color circle, the blue stands opposite to an orange; upon mixing, these colors give gray. In Ostwald's color circle, the blue stands opposite to a yellow, the pigmentary mixture yielding green. This fundamental difference in construction means that Ostwald's color circle is not serviceable to painting and the applied arts.

Experiments

I then set about understanding what was 'right' or useful in the artist color wheel, rather than dismissing artists as merely obdurate and stubborn. What I noticed primarily was differences in value between the same hues in the RGB and RYB systems. I next measured the hues and values of the scanned color wheels and entered the hues and values on a spread sheet. These measurements were used to generate radar (circular) charts of the hues and values. The hues, measured in degrees in the RGB color picker, were plotted on the diameter of the chart while the values were plotted as distance from the center.

In May of 1995 I posted the results of this study on the World Wide Web as "Computer Color and Artist's Color" I listed four general differences between the RYB and RGB system. The most important demonstration in this study was a figure showing the RYB and RGB color wheels transformed to gray values. In the RYB system, green and red are equal in value. There is an even gradation of values from yellow as the lightest at the top to purple as the darkest at the bottom. Left and right halves of the wheel are almost identical in value. The RGB color wheel shows no such symmetry or evenness in value, green of course being much lighter than its opposite red in RGB. I concluded that the artist's color wheel was skewed primarily to reduce the difference in value in opposite colors. Understanding what was correct and useful in the RYB system still did not create a tool for managing the larger gamut of colors available in RGB, nor did it create an RGB color wheel with harmonious opposites and triads. I then tried to reinvent the color wheel by various combinations of gradations and layers on the computer. I created radial gradations of red, green, blue, cyan, magenta and yellow in which the diameter corresponded to the value of the color and the placement of the resulting circle was determined by the value on the Y axis and the hue on the X axis. I was hoping that the overlapping of these circles would produce a complete range of possible hues and values in a useful juxtaposition.

Conceptually, a spectrum of all the hues of RGB arrayed on the X axis combined with a gradation from black to white on the Y axis should produce all the possible hue/value combinations. I discovered that the pasting and layering algorithms available on the computer did not produce even gradations or complete sets of intermediate colors. This led to a long series of experiments with layering, pasting, and opacity algorithms.

One of the algorithms I used produced a dominant peak of green in relation to red and blue and distorted the spectrum into what looked like a mountain range. I initially thought this algorithm distorted the hue/value relationship of RGB. In fact it was the value of the RGB colors that were distorted, and this chart put the hues back into their correct value relationship.

Now, instead of having a system that would reconcile artist's and RGB systems, I had found a system alien and antithetical to both. The chart of the hue/value relationship that I produced further showed that green comprised more than half of the total brightness of RGB and that white was therefor the highest value of green.

My chart of hue/value relationship of color also corresponds to the dominance of green in the spectral luminosity curve in Richard L. Gregory "Eye and Brain" Fig 6.4 and 6.5. Of this he says "The luminosity curve tells us nothing much about colour vision." because animals without color vision show a similar luminosity curve. P.94. Luminosity is discussed in the chapter on brightness and not mentioned again in the chapter on color. This to me is like the Wizard of Oz saying "Pay no attention to that man behind the curtain."

Results

The greyscale conversion algorithm is a useful tool because it allows a dramatic visualization of the relation of hue to value. Red, green and blue must be balanced on computer and video screens to produce white and shades of gray. Because the hue/value relationship is empirically derived from measurements of human perception and carefully adjusted and calibrated, it follows that information about human perception of color is imbedded in the RGB system for it to work.

Conclusion

I would like to propose the usefulness of "White is Green" in explaining the hue/value relationship of color. RGB color space is a distortion of the hue/value relationship in the same way that Mercator's projection map of the globe onto a rectangle is a distortion. Every pure hue in RGB is defined as having a brightness of 100% when in fact every hue has a different brightness. The artist's red yellow and blue color wheel creates harmonious colors only in a limited range of hues and values. Artists argue that harmony is subjective and culturally determined. Even so, once a convention or style has been established, only very specific colors will fit within that convention. Matching colors and creating variations within established limits requires careful control of hue and value, a task that I hope to make easier.

References