Functional Color

The Emir of Bukhara, reconstructed from the Prokudin-Gorskii 3-color photographic plates, around 1910

The photographs of Sergei Mikhailovich Prokudin-Gorskii (1863-1944) offer a vivid portrait of a lost world–the Russian Empire on the eve of World War I and the coming revolution. Taken between 1909 and 1912, the photographer exposed three glass plates in quick succession, each through a different color filter (red, green and blue). A full color image was recreated by projecting colored light through the plates, using the corresponding filters. The US Library of Congress has recently released a set of digital color images created from these archives that are stunning, with brilliant colors and fine detail. The archive is online, and includes a description of the process they used to recreate the images. This involves first scanning the plates to create three grayscale images, carefully aligning them, then assigning each separation to the corresponding color channel of an RGB image. As part of the process, they repaired flaws in the separations, then adjusted the resulting image to create “the proper contrast, appropriate highlight and shadow detail, and optimal color balance”

While very much admiring their work, my color-geek self immediately wonders how much these digital images really look like the ones the photographer showed his original audiences? The popular press and many internet viewers have marveled at their brilliance. But how much of this was visible in 1910?

Note the rainbow colors in the water, caused by the surface changing during the photographic process. Three-color lantern projector, Thomas Cradock Hepworth, 1889

Looking at one of these images on a digital display is equivalent to projecting with red, green and blue lights the same colors as the display pixels. This is how an LCD display works–you look at the light passing through red, green and blue colored filters. But, the filters used by Prokudin-Gorskii  are not the same as those used modern displays.

In an LCD display, the filter colors can be made very saturated and dark because we are looking directly at the light they produce. CRT and plasmas displays use phosphors, producing even more saturated colors, especially blue. Using a projector, the light passes through the filters and is reflected off of a screen or other surface. The result is invariably less saturated due to scattering, and also because a brighter light is needed, which means using less saturated color filters. While wonderful for their time, the audiences of the early 1900’s surely saw images that were much less vivid and colorful.

Rather than historical recreations, these images are marvelous visualizations of the information captured by Prokudin-Gorskii on his glass plates. I suspect he would feel they improve on the visualizations (that is, the projected images) he was able to create.  Appreciate them for what they are, but do not take their colors too seriously as a historical record.

Plato V terminals, Computer History Museum

Back when I was in college, I had the good fortune to become involved in the Plato IV computer-aided instruction project. The Plato IV system consisted of up to a 1000 terminals connected to a CDC mainframe computer. Each terminal had a black and orange plasma display, using the same basic technology now in plasma televisions. The 512 x 512 randomly addressable display, touch panel and typewriter-like keyboard created one of the first graphics workstations. (The Plato V terminals shown in the picture have the same technology. The Plato IV terminal wasn’t working when I was there)

The Plato system was used to teach courses during the day. At night, developers created courseware and expanded the system. The first computer chat, shared notes, and multi-player games were created for the Plato IV system.

My work involved connecting a plasma display, keyboard and touch panel to a PDP/11 mini-computer to create an “intelligent terminal.”   My thesis was on character encoding, to improve performance over the 1200bps communication line. I may have sent the first embedded graphics via email by writing software that packed an image into ASCII bytes that were then decoded in the terminal.

The plasma panel’s bright orange dots are created by neon gas discharge, revealing the fundamental mechanism of plasma displays. Plasma televisions use xenon to energize colored phosphors, but the Plato IV displays predate their development by a decade or more. While initially startling, both developers and users quickly adapted, focusing on the concepts presented, not the color.

At the Plato reunion, the question was asked whether having the orange and black displays was limiting. Would color displays have been significantly better?  As in all new media, creative minds work around and with the limitations of the medium. Creativity and innovation flowed in orange and black in the Plato project. Applications that required color were not explored, but there were plenty of other choices.

Would color have been better? This can’t be asked independently of the available technology. Given that there were no affordable color displays, it was clearly better not to wait for color. As display technologies evolve, adding color generally means lowering resolution. For text and illustration, higher resolution is the better choice.  Similarly, lower price and higher reliability are often more important than color.

All imaging technologies progress from a limited proof-of-concept (usually black and white), through grayscale to full color. Once the color versions achieve quality and reliability at a sufficiently low price, they will replace their achromatic  predecessors except for specialized applications. The history of  photography, printing, television and displays all show this evolution. The orange and black plasma panels of the Plato project of the 7o’s were the first step on the path towards the color plasma televisions available today. But more importantly, they enabled a world-changing burst of creativity and innovation whose impact can still be seen today.

Links: Plato History, Plasma Displays

“Color blindness,” or more formally, color vision deficiencies (CVD) affect roughly 10% of the population, mostly men.  Few people are truly blind to color; most simply have some degree of weakness perceiving certain parts of the spectrum.  People with CVD often compensate by being more sensitive to differences in luminance, texture, and position.

Predicting the effect of CVD is difficult because it affects color vision very early in the perceptual processing. In the eye, colored light is converted to three neural stimuli by the cones, specialized cells that live in the retina. Humans have three different cones, which are sensitive to different, overlapping ranges of wavelengths. Protan, deutran and tritan refer to people with anomalies in the cones most sensitive to long, medium and short wavelengths respectively.

Roughly speaking, protans and deutrans both have difficulty discriminating red from green, but tritans have difficulty discriminating between blue-purple and yellow. But, the reality is much more complex.

Many people with CVD can see colors when they are very bright and saturated, but not when the hues are more subtle, as in pastels or “earth tones.” A common problem for protans, who have difficulty seeing red light, is distinguishing brown from gray.  Deutrans have difficulty distinguishing green from amber, especially on displays, where the green is very yellowish. Both protans and deutrans have difficulty distinguishing blue from purple (which is blue plus red).

Fortunately, there are tools for simulating color deficient vision, such as Vischeck. The simulation models extreme cases of CVD where one cone is completely missing. However, any design that is legible in this extreme case should also work for people with less severe CVD.

Recommended reading: Colblindor, Color blindness viewed through Colorblind Eyes

Edward Tufte, in his classic book Envisioning Information, begins his chapter on color with the admonition: “Above all, do not harm.” Color used well can enhance and beautify, but color used poorly can be worse than no color at all.

For example, if text is displayed on a background that varies in lightness or color (like all too many PowerPoint templates suggest) its prominence and legibility will shift with its placement. Here’s an extreme example, where both the text and the background colors are changing. Some words pop out, some are nearly unreadable. Looking at the luminance view we can see why words are disappearing; insufficient luminance contrast.

Changing the text to yellow makes it all readable, but the color shifts with the background. This is the result of simultaneous contrast; the color of the background changes the color of the foreground.

Changing the text to white eliminates this problem, but there is still contrast variation. And, the colorful background is commanding more attention than the text. Which is more important?

Best to keep the colors simple so your message speaks clearly. When in doubt, stick to neutral colors, or even just black and white.

Consider this map of Point Reyes and its surrounding area, created by the National Park Service (PDF). Like most well-designed maps, it is an excellent example of functional color. Color is used systematically to label the different regions (water, land, parkland), and to indicate the different types of roads (major highway, secondary highway, surface roads). The colors are clear and distinct, but not gaudy. Colors are chosen according to simple, well-established conventions: blue for water, green for parks, red for highways. But, the color choices within these conventions have much to teach us about the design and perception of functional color.

The highways look to be drawn in two shades of red. In fact, the dark red highways are the same red as the lighter ones. They are outlined in black to make the color look darker, a trick of perception exploited by map makers to save on printing costs. Back when map printing was a fine art, each different color required its own printing plate and custom ink. Using a black outline to create a new red saves the cost of adding a new color plate. The black outline also makes the edge of the road stand out more clearly from its background.

The perceived difference between two colors is called “contrast.” Contrast defines legibility, readability, and directs attention. Designers and vision scientists describe colors in terms of hue (red, blue, purple, etc.) and lightness, plus a third dimension variously called saturation, chroma, or colorfulness. The red roads contrast in both hue and lightness with the light colored backgrounds. The network of red highways creates the foremost visual layer in this map, as this map is designed for drivers. The roads are labeled in black, which maximizes readability.

The text that labels the features in the light blue water is a slightly darker, more saturated shade of blue. Because our visual system groups objects that are a similar hue, this creates a “blue” background layer that is quietly legible yet unobtrusive. Similarly, Point Reyes is labeled in green to associate it with the green park region, but more emphatically, as it is the destination for this map.

All colors have a perceived lightness (luminance) independent of their hue. Luminance can be measured, captured by technology such as black and white cameras, or computed. Consider our map in a luminance view, which was created by the “Grayscale” function in Adobe Photoshop. The first thing to notice is that the map is completely readable and usable. The roads are still the dominant visual layer because they contrast strongly with the light gray land. The difference between Point Reyes and its surrounding ocean is more subtle than in color, but still visible. It is now easy to see that the text labeling Point Reyes is darker (higher contrast) than the text “Pacific Ocean,” which gives it more emphasis.

Reducing colors to their luminance values allows us to evaluate the contrast and spatial relationships in a design without the distraction of hue. Luminance contrast alone defines shape and edges. Metrics for text legibility are defined with respect to luminance contrast. Differences in contrast direct attention and define importance. Applying color cannot repair a design that is poorly organized or lacks a clear information hierarchy encoded by luminance contrast.

Designers in many fields call this “get it right in black and white.”

Functional color is designed for a purpose, color whose aesthetics are linked to its application. Now, it could be argued that all color has a purpose. But here, I will be talking primarily about color used in presenting information, such as a slide presentation, an illustration, or a data-driven visualization. Functional color choices are legible, appropriately direct attention, and create results that are both informative and aesthetic.

While it could be argued that any good visual designer can pick effective colors for a specific application, an appalling number of “professionally designed” displays use color combinations that are either illegible, garishly distracting, or both. (Slide templates are rich with examples). One goal for this blog is to present some basic principles and examples, in hopes of reducing the “color chaos” that so badly obscures important information.  But more significantly, I want to discuss why, including the technical, visual and perceptual principles that make functional color effective.

I’ve done a lot of teaching and presentations, and have even written a book, A Field Guide to Digital Color.  With this blog, I’m hoping to create an on-going forum for topics ranging from perception to aesthetics, with emphasis on those that either create or explain functional color.