This article is a "how to" about color calibration of astronomical artwork. I want to talk about calibration of the dark and the black: Estimates of black portions of the image using a digital single lens-reflex camera, hereafter called DSLR, to explore a perfect black portion of the night sky. This is usually the first step to obtain any astronomical photography.

Note: Click on the images to quickly compare images.


Color of stars is important to both, professional astronomers, and amateur astrophotographers. Sometimes it is also discussed, what is the true color of an astronomical target in a picture. Astro-artists probably have a different gut feeling about the proper coloring of their astronomical images. So from whatever perspective we talk about coloring astronomical pictures, expectations and results might be different.

The special case of a digital color camera, like a DSLR, is somewhat different from artwork in professional astronomy. These kinds of cameras were originally designed to match the visual color perception of the human eye as good as possible. Obviously, digital cameras will be the best choice to show perfect true color of astronomical objects. If there wouldn't be a couple of problems with these cameras: Their sensitivity to red light, especially the red hydrogen line, is quite low. Long exposure imaging will create some bias from thermal electrons. The night sky is not black. Just to mention a few of our problems. This is why astronomers, like me, use modified DSLRs. These modified DSLRs typically have the factory filters removed and replaced by different ones, that provide more light in the red and blue range of the visual spectrum. Why do we use modified cameras? That's easy to answer: You not only want to see a star cluster, but also a faint red gaseous nebulae, like the Orion nebulae or North America Nebulae, right? If your camera looses 80% of the red light, where the interstellar gas emits the most, your exposure time will increase by factors. Exposure time can reach a few hours to record an image of the North America Nebula. With a modified camera you can do a 1 minute exposure to detect the nebula and 16 minutes to get a reasonable result to be shared with your friends on facebook.

Dark currents

Astronomical imagery is different from ordinary photography. As already mentioned briefly, a silicon device to obtain pretty pictures will behave in a strange way. It will produce some "spooky noise" from long exposure, even if we close the optics and let the camera record the black. Artifacts of such long-exposures are in the range of dark currents, amplifier "glow" or other electronic bias added to the image just by taking a picture within minutes of exposure time with no light.

To calibrate such artifacts, there is a standard calibration procedure in astronomy, that has been already evaluated in the 1980ies of the last century. A very first article about this standard procedure has been published by Craig Mackay (1986). In year 2007, I claimed, that CMOS devices will override the CCD cameras soon. I decided not to invest into CCD any longer - and actually it happened: The modern CMOS sensors are so much better in astronomy, than the good old CCD guy, that they completely changed the game until today. The good thing is: Although silicon imaging detectors changed a lot since the time of writing of this paper, it is still true and proven by literally thousands of professional and amateur astronomers. And Craig Mackay was a wise astronomer, already mentioning CMOS detectors at a time, even though they performed bad in the 1980ies. So we can be pretty sure and still apply the techniques to our modern artwork to improve astronomical artwork. 

Dark current is any signal bias of a silicon imaging device, that will increase with temperature and exposure time. Dark current is caused by freed electrons, that are "kicked out" from the silicon crystal by thermal movement of the vibrating atoms at higher temperature, like environmental temperatures. The free electrons are not caused by illumination of the imaging sensor. But, an electron is an electron, and we cannot distinguish which one is freed from thermal signal and which one is freed by exposure to light. To reduce dark current, sensors will usually be cooled. For a usual DSLR cooling is difficult and works best in winter season. But we can apply a trick to reduce thermal signal. Dark current is calibrated by taking one or more pictures without light with exact same ISO settings and same exposure time as the image of the night sky. The number of dark images shall be the same as the number of images used for co-adding, if co-adding is used to combine multiple images into one. From years of experience using modern CMOS sensors in astronomy, I found, they differ a bit from the good old CCD guys. Therefore, it shall be noted, that dark exposures are typically recorded at the beginning and end of our session. This to obtain best results and match the same temperature range of our nightly session. I should note, that keeping libraries of dark signals is not a good idea, because hot pixels in modern CMOS sensors will(!) vary and change over time. Dark currents are then simply calibrated by subtracting a "master" dark from our image. This way we remove a statistically evaluated amount of electrons measured from our image, that are built up from thermal signal only, by measuring a thermal signal as reference. A master dark is an average from all recorded dark exposures. Calibration of the dark current is typically done by a dedicated astronomical software, like ArgusPro SE, that I developed years ago to explore the limits of DSLR astronomy. ArgusPro SE is a modern successor of a program called "Argus@Pro". Argus@Pro has been developed by myself in the early 1990ies. It was a very successful program at that time, the first astronomical program ever coming with a complete graphical user interface to control cameras, record images at fully automated image processing pipeline and photometry with a few mouse clicks. It was also used at the 1.54m telescope at the European Southern Observatory in Chile to perform speckle interferometry of faint stellar clusters and close binary stars.


Figure 1: Result from using the dark cloud I close to NGC 7000 as a black point estimated looks promising, but resulted in a loss of the very faint structures in the field of view. This cloud contains many faint stars. Even if the stars aren't really visible (because the sky background is not properly determined) they add reasonable bias. Ths black point estimate is too high resulting in loss of nebula contents and faint stars.

Figure 1: Result from using the dark cloud I close to NGC 7000 as a black point estimated looks promising, but resulted in a loss of the very faint structures in the field of view. This cloud contains many faint stars. Even if the stars aren't really visible (because the sky background is not properly determined) they add reasonable bias. Ths black point estimate is too high resulting in loss of nebula contents and faint stars.


Vanilla Skies

Removing any thermal bias from our images by subtracting the dark current was just a first step. The next problem is another kind of "black" signal. It is caused by the night sky. Independent from where you are trying to take pictures of the night sky, it will never will be black. In former times, astronomers reported a glow of light from long-exposure imaging, that is caused by the ionised atmosphere of our planet earth. After sunset ionization and re-combinations of atoms in the outer atmosphere will lead to a faint glow, that has been reported at different orders of stellar magnitudes for different astronomical sites. Nowadays, the human species adds higher orders of illumination to the faint glow of the atmosphere. This is caused by illumination of our cities and industrial sites at night.

How to cope with light pollution and atmospheric glow? Again, the answer is simple: We again remove it, but we need to properly fit the additional "dark" signal from our image. This procedure can be quite cumbersome, as I will illustrate here. For further discussion here, I use an image of the object number 7000 from the New General Catalogue of nebulae, NGC 7000, also called the North America Nebula. The nebula is a well-known extended H-II emission region in Cygnus star formation. My image has already been featured on my website. So we will play around with it for now.

All we need is an astronomical software tool, that can easily handle measures in the image. In astronomy science we call these software tools photometric software. We will use a simple aperture photometry for the demonstration here. Aperture photometry of a star typically uses all light from the star and also tries to fit the background around the star to obtain the portion of light, that comes from the star only. The image of a star, however, is never prefect. Usually a star appears somewhat blurred in our image. All we need to do is measure a black portion of the night sky to subtract artificial illumination from the earth grounds and air glow. Here it comes, why I choose this object type: It is very difficult to find the right portion of the night sky in the neighborhood of NGC 7000. So it will not be easy to obtain any perfect measure of artificial illumination to be removed.

Calibrate Black Point

16 images of NGC 7000 were taken with a modified Canon EOD 60D, Astronomik UV/IR block (L) filter and Canon EF 100 mm macro tele optics. Settings were put to ISO 400 and 1 minute of exposure time. The images were added together to obtain a single co-added image. If you want know, why ISO 400 is sufficient, you may find the answer on my website elsewhere. The images are corrected for dark current and flat field to also avoid vignetting from our 100 mm tele optics. If we stretch image intensities from highest intensities found (stars) to zero level, the result is a dark image with few stars. The image intensities have been then stretched a bit more to find the North America Nebula in the image.

The whole field around NGC 7000 is filled with extended H-II gas and also crowded with many, many stars. To find a reasonable black point calibration area to determine sky background, we use the dark clouds in this area. We select a rectangular aperture to measure pixel intensities. The first image shows the result from a measure of black in a dark cloud closely located to the nebula. The result already looks very promising. But, we try another dark cloud area, as well. The second image reveals, there are even fainter features found in the image, if we select a cloud, that is more dense and showing less faint stars in the field. Lets call it "more black" in this case. Hence, the second image provides more and fainter stars and reveals even fainter content of the nebula. A slight color change is also found between the figures 1 and 2, because the background is slightly different in colors. That is important to the color of the nebula itself, as its intensities are low. That is interesting to note: Colors of the very faint objects depend heavily from proper calibration of the sky background! Hence, we trust more in the second dark cloud measurement of "more black". We will take this cloud measurement for a reasonable black point estimate.


Figure 2: Result from using the dark cloud close II to NGC 7000 revealed fainter structures of the nebula and the surrounding faint stars in the field. The measured RGB values are lower compared to the previous measures from dark cloud I. The image appears brighter overall, because the sky background found in this cloud was found lower. Also note, there is a slight difference in color appearance when comparing this image with the previous one.

Figure 2: Result from using the dark cloud II close to NGC 7000 revealed fainter structures of the nebula and the surrounding faint stars in the field. The measured RGB values are lower compared to the previous measures from dark cloud I. The image appears brighter, because the sky background found in this cloud was found lower. Also note, there is a slight difference in color appearance when comparing this image with the previous one. This is because the blue bias did not change much in contrast to the red and green values.


With astronomical imaging, black comes in different flavors. Bias signals shadowing the black signal level need to be calibrated well. The most important bias signals to black are (1) dark current from thermal electrons and (2) non-black portions of the bright night sky, that are caused by artificial illumination and from glow of the ionised atmosphere of the earth. If calibration of dark signal and flat field is done properly, then imagery is prepared for the next step: Calibration of true colors. The Cygnus star formation is crowded with stars. Selection of a black portion of the sky will not be an easy task in this case. Even the faint invisible stars will add enough light to the night sky. In this case, the sky background can be determined wrong. Wrong dark level in the image will mean to loose faint structures of the nebula and background stars. Color of the very faint objects may also be affected by false determination of the black level. Hence, proper selection and measurement of the sky background is an important pre-requisite to color calibration.

In my next article, I will talk about the White Star Problem.



Mackay, Craig D., 1986: Charge-coupled devices in astronomy. Annual Rev. Astron. Astrophys., Vol. 24, p. 255-283, External link