True Color

previous | first

I have read passionate discussions about the true colors of objects that are captured by astrophotographers. I of course have some of my own opinons on this matter, but they are actually fairly gentle.

I like to advise people to consider "the color of a tree falling in the forest with no one to see it." Without the brain of an observer, there is no color, only light (and even that statement is suspect). Since the light levels involved in viewing distant galaxies and nebulas are well below the threshold of the color cells in our retina, it is really only speculation when we discuss what color we would see if we could see color. This is like asking "what color is infrared light if we could see it?"

On the other hand, we can ask the question of what would we see if we could somehow amplify the light to photopic vision levels. One might expect that a telescope could provide this amplification, but based on my viewing experiences, I suspect that other factors are at work in limiting the sensation of color at the eyepiece. It is an open question for me whether given a large enough telescope, I would see the Crab Nebula for example, in the colors as it is often presented in photographs. I don't know how the Great Red Spot on Jupiter got its name!

Certainly the methods to predict these colors exist, but many of the discussions on astrophotographic color don't use them, or use them incorrectly. To represent the color of an object it is necessary to apply the human visual sensitivities to the spectrum of that object. To render a color that matches it, the characteristics of the display (or printer) are needed.

As an example, let's calculate the RGB levels needed to render the color of an emission nebula, a gas cloud that emits light at a single wavelength, the hydrogen-alpha line at 656.3 nm, using a CRT monitor. The color is obtained by evaluating the tristimulus values for the human visual system. Because this is a single spectral line, the weighted integrals that are usually required are simplified to a lookup of sensitivity functions at that wavelength. They are (as tabulated at 655nm):

X = 0.2187,
Y = 0.0816,
Z = 0.0

To make a color that matches this perfectly, one would need a display that uses the hydrogen emission line as one of its primaries. Maybe such a display exists somewhere for some purpose, but it is not practical as a common output device, and wouldn't solve the general problem anyway: it is not possible to reproduce pure spectral colors, without using pure spectral sources. Displays and printed images use highly saturated, but not pure, primaries to reproduce the colors that fall between them, and they are unable to generate the colors beyond. Those in the reproducible range are said to be within the device's color gamut.

This becomes evident when we do the next step for rendering this color, converting it from tristimulus XYZ to display RGB. The operation uses a transform matrix that represents the phosphor colors of the CRT. Using the transform for the Rec. 709 digital video standard phosphor set, we obtain

R = .583
G = -0.059
B = -0.004

The presence of the negative amounts for green and blue is the giveaway that this color cannot be rendered using this display. Other colors might generate values greater than 1, also indicating that it is "out of gamut". When rendered as a print, the distance outside of gamut is even further.

The best we can do is select a color that we can make, to substitute for it. This is called gamut mapping and is an active topic of current color research. The results so far indicate that the best choice for a substitute color depends on both the behavior of the output device and the purpose of the image! If one does the obvious thing of clipping the negative values above to zero, the representation of the emission nebula is in shades of device-red only, probably a good answer for this application.

What about all the colors that are inside of the display gamut? While it is true that we can determine a mapping between an object's spectrum and a display value that matches it, this calculation depends on providing certain viewing assumptions. These assumptions are rather arbitrary in the context of viewing an astronomical object: a person's whitepoint adaptation state is one of them. Is it really meaningful to assume that the viewing environment is similar to noontime daylight?

The point of this exercise is to illustrate that "true color" is a term that really only applies in rather limited situations. The red colors of nebulas cannot be reproduced on displays, film, or print. The colors that are substituted for it will be different on each output device. The object colors that can be reproduced are not necessarily the same as their spectral appearance, even in a virtual telescope. Knowing this, astrophotographers should concentrate on "relative color" and bring out the detail and beauty of their images without excess concern over its absolute color fidelity.