Mixing Colors Digitally - Introduction

What facet of traditional art is most difficult for painting software to mimic? It's ‘mixing colors’ with physical pigments. For that reason, digital artists must totally rethink conventional color mixing recipes, suggestions for limited palettes, and strategies for color harmony in a digital context.

Here’s one example of the difficulty. The colors #1B4CD1 and #FDEC23 above, which resemble Daniel Smith watercolors Cerulean Chromium and Hansa Yellow Medium respectively, mix digitally to a yellow gray. It’s a pleasant unsaturated color, but not one of the lovely greens we expect from those pigments. What’s going on?

To understand ‘mixing colors’ digitally, we must go back to the proverbial drawing board.

Mixing vs. Blending Colors Digitally

For the sake of clarity, let me stipulate a distinction between mixing and blending colors digitally. In both cases, of course, we’re doing things in the software that require it to adjust the pixels (by variously illumining their red, green, and blue-emitting phosphors) that stimulate us to perceive colors on digital displays. But that’s a mouthful. So, to avoid saying it over and over I’ll just speak in shorthand, as we commonly do, about mixing colors and blending colors digitally.

By “mixing colors” I mean using software processes that produce a new color from the original colors you’ve combined on a single layer. There are ways to emulate dry mixing techniques (in oil, acrylic, watercolor, or pastel media): you apply colors to a single layer and then ‘shove’ them around with a loaded or unloaded paint brush, or a blender brush. And there are ways to emulate wet mixing techniques: you apply colors (and, perhaps, ‘clear water’) to a single layer and rapturously watch as they appear to flow together. In each case, the software employs sophisticated algorithms to order those colors around the screen based on factors like the original colors, where they are placed, how much is applied, the ‘wetness’ and ‘texture’ of the canvas, the real media being emulated, and so on and on. These amazing algorithms are not only labyrinthine, they’re proprietary and top-secret. In other words, I have no idea how they work and won’t be talking about them. But once the software has ordered a pixel to show so much of this color and so much of that one, how that new color is determined is straightforward, and that’s what I’ll be explaining. I’ll call these new colors the “mixed colors” and refer to the digital process that produces them as “mixing,” because the software is trying to mimic how painters mix colors with physical pigments.

By “blending colors” I mean adjusting the blending modes of layers on which you’ve already applied colors. This can mimic a range of traditional techniques like glazing color over dry paint, optically mixing soft pastels or colored pencils, and so on. I’ll call the new colors “blended colors” and refer to the adjustments that produce them as “blending” because the software already uses the term “blending modes” for the optional ways that layers can interact to produce color.

The Complexity of Mixing Colors with Physical Pigments

The main reason painting software can’t mimic how painters mix colors with physical pigments is this: the multiple physical processes involved in mixing pigments are too complex and quirky. Digital processes, by contrast, are quite regular and simple. They need to be, because there’s only so much processing power available, unless you’re painting on a supercomputer.

Let’s start with some complexity of mixing colors with physical pigments. Pigments are composed of granules, in varying shapes and sizes, that absorb, reflect, and diffract incident light in varying degrees. These granules are carried about in a binder that also influences the passage of incident light. The so-called “opaque” pigments have chunkier granules more likely to clump together than “transparent” ones.

When you mix two pigments, their granules become jumbled around. Most photons of incident light bounce crazily from one sort of granule to the other, like the ball in a pin-ball machine, before bouncing back out toward the viewer. Each sort of granule absorbs some wavelengths of the photon’s spectral energy distribution, which is why this is often called “subtractive” mixing: certain parts of the photon’s spectral energy are ‘subtracted’ away by each sort of granule.

But there’s more complexity. Some photons of incident light strike just one sort of granule, so their energy spectrum loses only one part or the other, depending on which granule they hit. When viewers ‘see’ these photons, they additively average the two colors they cause (and the third color caused by the photons that underwent subtractive mixing), much like they do the tiny dots in a pointillist painting or the variously painted regions on a spinning top. This additive-averaging psychological process and the subtractive mixing process yield different perceptual results: they don’t reinforce each other.

A third level of complexity is introduced by “color metamerism,” which is when the same perceived color is produced by different patterns of spectral reflectance. For instance, two pigments can look the same color in sunlight, but different colors in incandescent or florescent lighting. Paint manufacturers aim to produce a certain color under studio lighting, but they might do this with different pigments. This implies that we cannot predict from its color exactly how a pigment will perform in a mixture. For instance, two blue pigments might be metameres: they look alike under studio lighting, but their granules absorb spectral energy differently from incident light, and thus the pigments ‘mix’ differently.

(For more on the complexity of mixing colors with physical pigments, see David Brigg’s Dimensions of Colour, “Part 6: Colour Mixing in Paints.” My summary above is indebted to his careful research.)

Mixing colors with identical physical pigments is also quirky. The size, shape, and concentration of granules will vary among brands, as will the binder. These factors influence how the granules absorb, reflect, and diffract photons, and how the granules become distributed through the mixture. It’s no wonder, then, that artists make color charts to track of how their paints behave in mixtures.

Given these complexities and contingencies, it’s clear why digital painting software can’t mimic how painters mix colors with pigments. So, how does software mix colors?

Mixing Colors Digitally 

Recall from an earlier post, “Thinking of Colors in RGB,” how painting software ‘thinks’ in RGB color space: it identifies and adjusts a color in terms of the illumination (on a scale from 0 to 255) of its monitor red, green, and blue channels. For instance, the color monitor red is (255, 0, 0) or #FF0000, and the most colorful yellow is (255, 255, 0) or #FFFF00.

How does the software determine which color to produce when we mix those colors on a single layer? There's a common misunderstanding that the colors mix additively. That’s based on a confusion between two stages of the process. At a later stage, after the software determines which color to produce at a given pixel, the illumination of red, green, and blue phosphors at that location will combine additively to produce the color. But the earlier stage for determining that color is not additive. If it were, this red and yellow would mix to the same bright yellow (255, 255, 0) or #FFFF00, because the total illumination of a component light cannot exceed 255. Indeed, digital artists could mix only a few garishly intense colors and everything else would be pastel tone or white.

(Painting software doesn't mix colors additively, but it can blend them additively. To do this, paint one color on a layer set to Normal blend and the other on a higher layer in the Lighten blend mode. The overlapping area will be the additive blend of the colors.)

Furthermore, which color to produce isn’t determined in either a subtractive or additive-average manner, the two processes that are prominent in mixing colors with physical pigments.

Rather, when you mix colors on a single layer, the software keeps the math very simple: it will weight average the red components, green components, and blue components of the colors. Sometimes the results are similar to what we expect from mixing the colors with physical pigments. For example, when these red and yellow colors are mixed digitally in even amounts, 50/50%, the color displayed is {[(50% of 255)+(50% of 255)], [(50% of 0)+(50% of 255)], [(50% of 0)+(50% of 0)]}, which is the bright orange called (255, 128, 0) or #FF8000. When the colors combine at a location in unequal proportion, the weighted average is used for each component. For instance, if there is 80% monitor red and 20% yellow, the calculation is:

Red component     = {(80% of 255)+(20% of 255)} = 255
Green component  = {(80% of 0)+(20% of 255)} = 51
Blue component    = {(80% of 0)+(20% of 0)} = 0

This mixed color (255, 51, 0) or #FF3300 is a bright, saturated orange red.

Such orange colors are very close to the colors we could mix with physical pigments.

Often, however, digital mixes don’t conform to traditional expectations. Recall the first example above. Now we can see what went wrong: when we digitally mix colors that resemble Hansa Yellow Medium (253, 236, 35) and Cerulean Chromium (27, 76, 209) in equal amounts, the weighted average is a yellow gray (140, 156, 122) rather than the green we expect.

Conclusion

Mixing colors with physical pigments is a complex and quirky process. It’s largely dominated by a physical subtractive process but involves a psychological additive-averaging process as well.

Mixing colors digitally, on the other hand, is a different, but simple process of weight-averaging.

So, how do the results of these mixing processes compare? I’ll share experiments that clarify some of the major differences in the next two posts, “Mixing Colors Digitally—Searching for Greens” and “Mixing Colors Digitally—Searching for Darks.” That’s a first step toward developing better guidelines for mixing colors, using a limited palette, and achieving color harmony in digital painting.

***

Thanks for reading!

I hope that you enjoyed this post and that it inspires you to enjoy digital painting. If you find this post helpful, please share it with your friends. And please send me your insights on digital painting and suggestions for Digital Paint Spot.

Bob Kruschwitz