The technological capture and reproduction of images dethroned the artist as magical image maker and robbed all pictures of their aura of rarity and preciousness, leaving us drawers and painters with the same status as those oddballs who insist on writing novels in longhand or doing all calculations with a slide rule. On the other hand, analog and digital imaging technology is a most amazing box of educational toys for learning about aspects of perception and light. I’ve had a long-running obsession to understand as much as I can about how these technologies work, from chemical color film to digital image processing, and studying and playing with these things has deeply informed the way I approach observational drawing and painting. In this post I’ll share some samples of such play and how I learn from it. I will try to make this both fun and informative – if I’m explaining stuff you already know, feel free to skim through.
As you probably recall from science class, Isaac Newton demonstrated that white light is a combination of all colors of light, and that the individual wavelengths of light appear to the eye as the different colors of the spectrum or rainbow. Red is at the long-wavelength end of the spectrum, and as the wavelengths get shorter, the color transitions to orange, yellow, green, cyan, blue and violet. Later experimenters discovered that a wide range of colors could be reproduced by combining just three colors of light, one representing the long (red) wavelengths, one the middle (green) wavelengths, and one the short (blue) wavelengths. The illustration below represents the overlapping beams from spotlights of these three colors. Where all three overlap, the light is white. Where red and green overlap, we get yellow. Blue and green make cyan (which you might call turquoise, aqua, or teal), and red and blue make magenta (or fuschia, reddish purple). With red plus green, but more red than green, you have orange, and so on. This kind of color process is called RGB, for the red, green and blue lights that are used.
This way of making colors by combining three colors of light in varying ratios is called additive color mixing. It’s the basis of color television, cathode ray tube screens, liquid crystal displays, video projectors, and the monitor on your smartphone. Here’s a close-up of an LCD computer monitor. A screen has thousands or millions of pixels (short for picture elements), and each pixel has a red, a green, and a blue element. A digital picture is nothing but a series of numbers representing the brightness levels of each of the three colors for every one of these pixels in a grid.
The photo below contains 305, 400 pixels, each one defined by levels of red, green and blue light specified by numbers from zero to sixty-four. This is a small version – the original camera photo had over ten million pixels.
With a digital picture it is easy to separate the three component colors as “channels”. If we take just the levels for the red component of each pixel and render those as a monochrome image, we get the result below. The skin looks light, almost luminous. Taking a photo with black and white film through a red filter would give a very similar effect. Most of the variations in skin tone are variations of redness, so when red is all you can see the differences are minimized.
Doing the same with the green channel gives a pretty objective black-and-white rendering of the original photograph. Because the green wavelengths are in the middle, or average, of the spectrum, they’re pretty close to the average lightness levels, without distortions in tone. The red channel made me look youthful and glowing, but the green channel shows my age a bit more objectively.
The blue channel is similar but the effect is even harsher. The skin looks darker and blemishes and discolorations of the skin are more pronounced. Some of the early black-and-white photography processes, including the film used for early silent movies, were sensitive only to the blue end of the spectrum, so they tended to render skin as dark and blotchy, necessitating the use of white make-up on the actors.
With this portrait photo, the red channel is strikingly different from the green and blue channels, which are more like each other. If I had used a landscape photograph for the demonstration, the blue channel would be the one that stood out, with black foliage and a stark white sky, while the red and green channels would be more alike.
There’s another kind of color reproduction, called subtractive color mixing. This is used in printing and in photographic prints or slides, where you start with a white ground (all wavelengths) and filter or absorb wavelengths selectively using dyes or pigments. Transparent paint, such as watercolor, is essentially a subtractive color mixing technique. The standard colors used in subtractive color processes are cyan (slightly greenish blue), magenta (purplish red) and yellow. As you can see from the illustration below, mixing all three colors doesn’t give a perfect black, so a fourth layer of black ink is added in four color process printing. This kind of color process is called CMYK, for cyan, magenta, yellow, and “key” (black). Note that the subtractive process uses as its basic colors the same colors that are the combined colors in the additive process, and that the combined colors in the subtractive process (the overlapping areas below) are very similar to the basic colors in the additive process.
Here’s an enlarged illustration of an image printed in a CMYK process. Where the RGB process varies the brightness of the colored elements, the subtractive process varies the size of the colored dots. In both types of image, you’re only seeing three colors, but they blend in the eye to create the illusion of a full range of colors.
The image below is from “Butterflies and Flowers”, a performance by Claire Elizabeth Barratt and her Cilla Vee Life Arts company (with whom I have occasionally collaborated) at the New York Botanical Garden in the Bronx in 2004. I chose this image to play with because it has such a range of vivid colors.
To prepare this photo for color printing we would make “color separations“, the cyan, magenta, yellow, and black layers that would be successively superimposed to make the full color image. Here is the cyan layer, followed by a version of the image with the other three layers (magenta, yellow, and black, without the cyan. Notice how the red and yellow colors both look white in the cyan image, and how different the faces look in the different colors.
The subtractive process uses inks to absorb certain colors of light. Cyan ink absorbs red light, and reflects blue and green light, so the cyan layer of the CMYK image is equivalent to the red channel of the RGB image, and shows a similar smoothing of skin tones. The magenta layer in CMYK corresponds to the green channel in RGB, and yellow in CMYK corresponds to blue in RGB.
I’ll do the same thing with the other layers, showing each single-ink layer followed by the full image minus that color. Here’s magenta and minus-magenta.
You’ll notice that the “minus one color versions” look like different types of faded images. Old motion picture film often loses its cyan layer, giving a reddish image like the “minus cyan” example three images up. Color inkjet prints that have been displayed in the sun often lose their magenta layer, leaving a greenish image like the one immediately above. Next, yellow and minus-yellow:
The black layer of a CMYK print is like a very light black and white version of the image. The lighter values will be distinguished by the colored inks, so the only place the black ink is needed is where the color mix doesn’t give enough contrast, in the darkest areas.
The CMY image without the black has the full range of colors but lacks contrast. It lacks a full range of lightness or luminance.
Aside from additive (RGB) and subtractive (CMYK) processes, there’s another way of digitally specifying the values of pixels using a different combination of variables. “Lab” color does not define color by the levels of light or pigment used to reproduce the color, though it still uses three dimensions. “Lab” isn’t short for “laboratory” – L (lightness), a, and b are the names of those three dimensions. The three scales are actually based on the way human color perception works in the brain.
The human eye has three different kinds of cones, or color-sensitive receptors, but interestingly, the peak spectral sensitivities of the cones do not correspond to red, green, and blue, but to something more like yellow-orange, yellow-green and blue. The visual cortex of the brain takes the input from these three sets of cones, and from the low-light sensitive rod cells, and, by comparing and contrasting, analyzes colors according to their variable positions on three scales: dark to light, reddish to greenish, and yellowish to bluish. That’s the basis of the Lab color model. It uses the numbers to define colors along these three polarities. In practice, the Lab color model is mostly used as an intermediary, to translate between additive and subtractive modes, but it’s a fascinating system to explore because it is such a good simulation of how the human visual system processes color.
When we translate our experimental image into the Lab color space, we can selectively “flatten” the channels, showing the image with one variable removed. Here’s the image with all variations in the lightness channel eliminated. All the color differences are here, but without differences in dark and light. It’s like the low-contrast CMY minus K version of the image (above), but instead of low-contrast, here we have absolutely no contrast in values, only in hue.
Now let’s restore the lightness channel to its full range and flatten the “a” (red/green) channel. The resulting image is very similar to simulations of the vision of people with complete red/green color-blindness. Deuteranopia or Protanopia are the most common forms of color-blindness, and also similar to the way dogs and cats see color. They have only two types of color-sensitive cones, so they can distinguish blue colors, but red and green colors all look more or less the same. Note that the red flowers here completely blend in with the green foliage background. There is speculation that the ability to distinguish red from green was evolutionarily advantageous because it helped locate fruit!
If we flatten the “b” (yellow/blue) channel, we can see the contrast between reds and greens but not between yellows and blues. Tritanopia, another, very rare, form of color-blindness, looks like this (below). For the person with normal color perception, the version below showing red/green distinctions is probably more pleasing than the version above that shows yellow-blue distinctions. The lightness scale can often stand in for the yellow/blue scale because we see yellow as light and blue as dark. The red/green scale is more equal in terms of values, but it is better at separating animals (usually reddish) from plants (usually greenish). The yellow/blue scale can be seen as separating land (yellowish) from sky and water (bluish).
In looking at the colors of subjects I am drawing or painting, I often try to understand them according to the Lab scales. Lightness/darkness is by far the most important scale to define form. Seeing colors on the relative “a/b” scales, as bluish vs yellowish and reddish vs greenish, is simple and clarifying. This model helps in observing subtle differences within color areas and help an artist avoid the “flatness” that often results when painters think of colors as duplicating surface colors of objects, rather than relative qualities of light.
Let’s try some other variations on this image, just for fun. Here is an “inverted” version of the full color image, essentially a color negative. Light becomes dark and dark light. Every color becomes its complement: blue becomes yellow and red becomes green.
Here’s a version with the Lab lightness channel inverted, and the “a” and “b” channels not inverted. Here the lights and darks are switched, but the hues of things remain the same as they are in the original image.
We can restore the L (lightness) channel to its correct orientation and instead invert the color channels. Here’s a version with the “a” (red/green) channel reversed. The dancers are green and the foliage is brown.
Here’s the “b” (yellow/blue) channel reversed. This makes the dancers’ skin look rather purple, and the foliage becomes blue. I find both of these variations psychedelically beautiful.
Finally, the image with both “a” and “b” channels inverted. In essence, this converts all hues to complementary hues while leaving values unchanged.
Let’s go back to the photographic self portrait and do some other digital manipulations on it. Here I have increased the contrast to separate only the brightest highlights of the image.
And next, I’ve increased the contrast to bring out only the darkest parts of the image.
Here I have combined the darks and the brights against a mid-toned background. This is essentially how I’m looking at my subject when I’m drawing with lights and darks on gray paper. The paper provides a mid-tone, and I draw highlights with white and shadows with black, getting a wide range of values much more quickly than would be possible by drawing with only darks on a white paper.
As I often do in drawing, here I’ve superimposed the colors over the simplified black/gray/white values to make a color portrait. As in the “Lab” model, the face is a little reddish, a little yellowish. Some of the background colors are a little bluish or greenish. Seeing color according to just three polarities simplifies it for the purposes of time-limited drawing. What I have done here with a digital deconstruction of a photograph is very similar to what I do mentally during the process of observational drawing or painting.