by Roger N. Clark
Color accuracy in astrophotography post processing can be difficult to maintain depending on the processing methodology. Here I investigate some processing methods that maintain good color versus steps that destroy natural color. A set of test files are present for you to test your own methods.
The Night Photography Series:
Tone Mapping at Raw Conversion Method
Challenge: Try Post Processing Yourself With These Images
About the Challenge
References and Further Reading
The challenge in this article is to produce an accurate color image in the presence of light pollution. Dealing with light pollution is a common problem in night sky and astro photography. For this article I use a commonly available color chart: the GretagMacBeth color chart. If you have one of these charts, and are viewing this article on a good color-calibrated monitor, you should be able to hold up you color chart and see colors that are very close to those in Figure 1. The bottom row of patches, between white to black should look like shades of gray without any color. The measured RGB values in the image in Figure 1 are very close for each patch.
For the astro-imaging test, "light pollution" was added to the scene by placing a grid of orange threads in front of the lens (Figure 2a). The threads were illuminated by sunlight. The signal from the threads is added light to the light coming from the chart. The threads blocked a tiny fraction of 1% of the light from the chart, so is inconsequential. This setup is an excellent analog for the light from an astronomical object with added light pollution in our night sky. Included too is the transmission of the terrestrial atmosphere because the illumination is the sun, which must pass through the atmosphere to illuminate the chart.
The light on the color chart and the threads were balanced so that the light pollution addition is similar to the astrophotography guideline of exposing so that the histogram for the sky with light pollution peaks about 1/4 to 1/3 of the way from the left to right sides of the histogram on the Camera LCD (Figure 2b).
The resulting image of the color chart with "light pollution" is shown in Figure 3. The challenge is to use a processing work flow that removes the light pollution and maintains accurate colors.
My methodology is simple and is described in detail in this series. Basically, it is the same color-managed work flow that I use for all my photography to produce reasonably natural color images. I use a raw converter, such as photoshop's, ACR), convert the raw file to a 16-bit tif image with Adobe RGB color profile, use lens profiles to correct for light fall-off, chromatic aberrations, and a small amount of noise reduction. I also use the black point slider to reduce the average offset from light pollution. See parts 3a, 3b, and 3c for more details and example photoshop ACR settings.
After performing the above raw conversion and bringing the image into photoshop, there was some additional light pollution gradient in the lower right corner. A quick feathered selection and subtraction with the curves tool in photoshop resulted in the corrected image in Figure 4a. Subtraction with the curves tool is simply moving the lower left point to the right, as shown in Figure 4b. The entire process from raw file to finished image took only a couple of minutes.
Compare the results in 4a with the colors in Figure 1. A smaller version of these two images are shown in Figure 4c. The results are very very close. This illustrates that color accuracy is maintained even when subtracting light pollution from tone-curve data, contrary to the dire prediction of internet "experts" who have proclaimed there would be horrible color shifts.
The internet is full of methods for post processing night sky and astrophotos. Most are not a color managed work flow, and many seriously warp colors to something not real. Here are some examples.
Common on the internet, as described in part 2d of this series, is to use a low Kelvin white balance temperature to reduce the effects of orange light pollution. White balance is a multiply, but light pollution is added light, so needs to be subtracted, not multiplied. Use of white balance to reduce effects of light pollution causes a bluing of the scene, especially at lower scene intensities (Figures 5, 6, 7). This produces very unnatural colors. As light pollution increases, the bluing effect becomes stronger.
Online I see many tutorials that say to align the histogram peak of the 3 color channels. That is WRONG. Aligning the histogram peaks suppresses the dominant color in an image. The colorchart has a variety of colors, so the problems aren't nearly as pronounced as in astrophotography, but we still see the detrimental effects. The image in Figure 5 included choosing a color temperature that gave the most neutral RGB values on the darkest square, and aligned the major histogram peaks. Clearly, the result is poor color. Steer clear of any tutorial that tells you to do this to your images! Avoid any form of histogram equalization (see also Figure 9) if you want natural color.
Compare the most accurate color image of the color chart in Figure 8 with the results in Figures 5, 6, 7, and 9.
Use the files below to reproduce what I did above, or try your own methods. If you use a traditional astrophotography linear work flow, I have included a bias frame and flat field. These are high signal-to-noise ratio images, so stacking is not needed. Stacking does not change intensities, it just improves signal-to-noise ratio. If your work flow needs multiple images and stacking, just copy first frame and increment the frame number, e.g. 8365 to 8366, 8367, 8368... etc. Can you produce accurate colors with your processing method(s)?
canon-7d2-mb-colorchart+lp-0J6A8365.CR2 image with "light pollution" 23 MByte file
canon-7d2-mb-colorchart+lp-0J6A8365.xmp photoshop ACR xmp file, setting for raw converter, 7.3 KByte file
canon-7d2-biasframe-0J6A8408.CR2 bias frame 19.5 MByte file
canon-7d2-flat-field-for-colorchart-0J6A8393.CR2 flat field 25.1 MByte file
If you do the challenge with the above images and post your results, you give permission for me to post a small version of your image here so I can discuss results of different methods.
The challenge is to check your work flow for color accuracy or present different work flows that can produce accurate colors.
Producing natural color images of the night sky is a challenge with the added light from light pollution. Color perception is complex, and a modern color-managed work flow can produce excellent results if light pollution (and other sky glow) is properly removed.
Obvious in the images produced here and in part 2d of this series, many post processing methods that produce bluing of the scene are all too common in the digital photography era as of this writing (circa 2000-2016). Natural color images of the night sky as of this writing are quite rare.
The internet "experts" continue their attacks. They insist that subtracting light pollution from tone-mapped data (like that produced by photoshop) produces unnatural color. Obviously, those charges are not true. Their ideas are based on a flawed understanding of tone curves and color spaces.
References and Further Reading
For linear processing of digital camera data, you must apply a color matrix conversion to get accurate colors. See this dpreview thread discussing the effects and how to appliy the matrix multiply.
Clarkvision.com Astrophoto Gallery.
Clarkvision.com Nightscapes Gallery.
The Night Photography Series:
First Published February 13, 2016
Last updated March 31, 2018