Milky Way Post Processing: Color Correction

I’m sure you did not skip the first two parts of this series, right? Did you? If so, please see Finding the Milky Way and Capturing the Milky Way. I’ll wait until you get back.

Back so soon? Hope you had fun reading about the Milky Way and how to photograph it. Here is a confession: You really do not need to jack your ISO up as far as I stated in Capturing the Milky Way. What happens when you set the ISO high is that you lose some dynamic range, and you will get some clipping (loss of highlights), and of course you increase the noise – BUT your processing will be a little easier because you won’t have to push any settings more than just a smidgen.

Hear are the general steps I take to attack my Milky Way images.

  • Noise Reduce
  • Color Correct
  • Contrast and local enhancements
  • Foreground/background blending

There are dozens of ways to do each of these tasks.  If you love Lightroom (I don’t particularly like it because it is SO slow to load and doesn’t allow me to blend multiple images) you will find some great resources by Ben Canales. For a $20 donation he’ll walk you step by step through his processing regimen.  The only downside to his tutorial is you must have web-access to view it – you can’t save a copy.

Even though I would normally noise reduce first, I am deferring the explanation for now and attacking the color balance problem. Sometimes all you need to properly color correct is to open your image in Adobe Camera Raw and use the White Balance Tool.

Much of the area near the Milky Way is “white” so clicking that diffuse glowing part with the white balance tool will properly balance your sky… or not depending on how bad the light pollution is.  Where exactly should you click? Not on individual stars (though that may work too if you pick the right colored star and you do not have clipping).  Just about anywhere except the brightest areas of the Milky Way should work.  It will not hurt at all to “click around” a bit until you get a natural look.  Here is a Milky Way image color corrected using the ACR white balance technique:

Milky Way Rest [C_049455]

However if the light pollution is pretty bad, you don’t have a raw file or your sky is quite orange/brown, you will want to employ a more potent solution.  This solution comes from Sky at Night Magazine.  Below is a video we recorded during our Photo Manipulation 150 Webinar.  One giveaway that your sky is not naturally colored is if it is orange, brown, green or completely blue.  I am not going to tell you not to render your sky like that – after all it is your photo and your taste will dictate what you want, but if you want people who enjoy astronomy to take your photo seriously do not go too far from reality.

One of my favorite images of the Milky Way resulted from allowing the camera to select a white balance. I used a blue-white LED flashlight and that caused the night sky to go “sepia”.  I did do some local enhancements to bring out the Milky Way. How I did the enhancement will be discussed in the next installment covering “Local Enhancement”.

Famous III  [C_035478]

If, however you want to get your sky naturally colored despite the light pollution, hopefully you’ll find this video informative – there are a bunch of additional tips, too!

You may have to enter the password BrownSky to watch it.


Related Articles Include

  1. Local Enhancement (Bump up Those Stars)
  2. Image Blending (Foreground O MaticEasy HDR)

Are we getting this right? Got a question? A quibble? Please leave a comment! And if this is really resonating with you, please share.  We love it when you share.

18 thoughts on “Milky Way Post Processing: Color Correction

  1. Marsha Kirschbaum

    Steven, when you say that we don’t need necessarily to shoot at the high ISOs you had previously mentioned, what do you mean by that? If I calculate the time needed to expose the stars based on the 500 / focal length that I desire and I come up with an exposure time, and I am already at 2.8, then my ISO is the only variable that I can raise to come close to that exposure time. Are you suggesting shooting at a lower ISO, and in post processing raising the exposure and eliminating the noise by stacking or some other post processing technique? Thanks.

    Also I get a “page not found” on the link to “skyatnight”

    1. Steven Christenson

      If you’re planning to use a 500mm focal length you’re in for some serious problems unless you use an equatorial mount. If you refer to my 600 Rule article you can calculate the amount of time a star will hover over a single pixel at 500mm. Only about 1/3 of a second! Since the star has moved on, exposing longer than 1/3 of a second won’t get more data. Please also refer to my “Exposing for Stars” article for an allegorical explanation.

      As for page not found, that’s too bad. Apparently they removed the page so I’ve removed my reference.

      1. Howard Brainen

        I’m pretty sure Marsha wasn’t asking about using a 500mm lens for Milky Way capture, but was referring to calculating maximum exposure time by dividing 500 by the focal length. We had learned this was a safer rule than the 600 rule since it results in a shorter exposure time, and therefore less streaking. However, in your excellent description of the 600 rule, where you calculate the actual time for a star to move from one “pixel” to the next, we now know neither rule even comes close to producing perfectly sharp stars.

        What we would both like to know is what you meant by: “You really do not need to jack your ISO up as far as I stated in Capturing the Milky Way.”

        1. Steven Christenson

          My bad. My response still applies… the 600 (or 500) rules are not at all aligned with the reality of the star movement. Once a star – and therefore a whole sky full of stars – has moved to a new place on the sensor no additional data is being collected except foreground (unmoving elements), sky glow and noise.

          Perhaps this analogy will help:

          Imagine you’re collecting water from a slowly leaking faucet in a graduated cylinder but the container is on a conveyer belt. Once the drip stops falling into the cylinder, there is nothing more to collect. You look at the scale on the cylinder and record the height to the nearest integer as your pixel value. You can increase the fine-ness or the coarseness of the markings on your graduated cylinder, but you are NOT changing the number of drips. At “unity gain” one drip more = one additional value. Below unity gain, 3 drips might be measured as a 1 – so you’ve lost information. Above unity gain, 3 drips might be measured as 9 – but you haven’t really gained any information because there are no 2, 4, 5, 7 or 8 values.

          Photoshop can be used to change 1, 2 and 3 to 3, 6 and 9 – you don’t need to make the camera do it. Since there is a maximum possible measurement we lose data if the gain (multiplier) is too high. Any value above a multiplier of 1 means we’ve cut the number of possible measurements. For example at twice the unity gain value, we can only measure half as much – because we are only keeping say two digits of measurement. At half the unity gain value two drips and three drips result in the same value – so we’ve lost information. In each case we are compromising the dynamic range. And finally, if our gain (multiplier) is too high we will exceed the maximum recordable value before we fill the cylinder, i.e. we get “clipping”.

          ISO at Unity gain is different for each camera/sensor and as we already observed, the focal length and sensor size dictate the maximum amount of time available to measure the photons from one star.

  2. Marsha

    500 mm? I wish! Though I am wait listed for the astrophotography 101 class. What I mean’t was 500 divided by 16 or whatever focal length I would be using for exposure time. I will re-read your 600 article. Thanks.

  3. Leandro

    Hi Steven, thanks for the information!

    Have you ever experienced pink skies when shooting with no moonlight? It happened to me here and here, both pictures were taken on the same area, in different nights with Daylight WB.

    Instead of getting an orange color, I got a pink one. Weird.

    I kept the sky that way because my main problem there is that I was using LED lights and if I take the WB down, it affects horribly the foreground. I know I could use masks, but it’s actually a motion controlled timelapse, so it’s not easy to edit.

    Anyway, what I’m actually asking is: why the hell is this happening? I mean: there’s little to zero light pollution, no moonlight; so why do I get that color? Maybe you know something I don’t.
    BTW, there was a lightning storm near by the day the first picture was taken, but not on the second one.

    Thanks again!

    1. Steven Christenson

      Great shots, Leandro. I’m sorry, however but I don’t have an explanation. I suspect dust or particulates in the air. I have also seen some programs (e.g. an HDR program) which had trouble mapping colors in RAW images. Those came out pink – but usually only when there was clipping. Good luck.

    1. Steven Christenson

      Thanks for asking, Mike. We’ve been busy here at SCA and failed to notice that we had already covered one of the topics in quite some detail: Image Blending. In the article “Foreground O Matic” we showed how to replace elements of the foreground. A similar article showed how to blend different images together in a very simple way (Easy HDR).

      Local Enhancement has been lightly covered in “Bump up Those Stars“. An almost done article, however, will reveal how the waterfall was brightened in this image (its not what you might think!), as well as how to reduce trailing of stars AFTER the fact. But one foreground brightening technique was covered in a 2010 article which has been revised and will be republished on November 20, 2012.

      Steven Christenson: New &emdash; Tumbler

      That leaves the “Noise” article which has taken on a life of its own.

  4. Todd Walker


    Thanks for all you have added to the page(s) here – it all really gives someone who doesn’t normally shoot at night a nice, ballpark place to start and learn…appreciate it a ton.

    In your last comment above, regarding “Foreground O Matic”, that tutorial itself is VERY good. But I’m curious on one thing; suppose the forground-blended shot of the satellite was not of a star trail, but rather a 3-5 exposure “stacked” image of the Milky Way. I’m not sure if what stacking-software used matters here, but I know aligning must happen first to line the stars for this stack. Unfortunatly this blurs the forground. Snapping a seperate, single exposure of the forground and then blending it with the stacked-stars would work i suppose, like you did with the satellite. But the forground/horizon in that shot isn’t that difficult to blend with. What blending-option would one have if let’s say the satellite were instead a tree, with many branches or other tedious objects to mask around?

    A simpler way to ask this question may be looking at your image above of the Milky-Way above the lake. Assuming this image utilized aligning/stacking several exposures for noise-reduction purposes, was it difficult for you to mask around all those tree tops?

    Thanks a lot for any response,

    1. Steven Christenson

      Todd, As I noted in the Foreground-O-Matic article there are multiple ways of blending including using a gradient mask. It’s a matter of how much work you want to do and how complicated interaction is between the sky and the foreground. Trees are perhaps a worst case for many reasons, including their nature and the fact that they tend to move.

      My “Falls under the Milky Way” shot is a single exposure using a high performing camera and less adjustment then you might expect.

      Also, as I noted under the “Astrophotography Technique to Conquer Noise” the best results for getting a Milky Way are when you have no foreground in the shot at all… however if you want to blend multiple sky shots together over a single foreground remember a few key points:

      1. For photoshop to autoalign the layers properly you have to mask off the foreground (auto align ignores areas that are covered by a mask).
      2. You need to anchor your alignment to match the foreground you intend to keep. And since the Milky Way will be either setting or rising, you’ll want the first or last image in your sequence as the anchor.
      3. Areas adjacent to the foreground may not blend well.
  5. Todd Walker

    Thanks Steven,

    My fault – I should have “re-read” your notes a little slower before posting my question. Your page states in exact words, “One alternative is to use a gradient mask. But that may not work either such as when a tree reaches up into the sky with lots of tiny branches. Another approach is to do hand masking/layering. The following provides a rough idea of both how to approach the hand masking problem and how much more difficult the process can become”. … that answers my question.

    I figured there’s no simple way around it, as masking a tree (or similar shape) just plain requires “work”. Seeing that I normally print at a mid to large size I would need to be careful in these areas.

    Here’s my first attempt at the Milky Way:

    Here I used DeepSkyStacker to align and stack the stars, then used a single exposure to mask/blend in the foreground. If an object appeared in foreground here it would take a little bit more work.

    Thanks You!

  6. Pingback: De Streaking | Star Circle Academy

  7. Graham Wiffen

    Hi Steven
    A great set of articles and very interesting to a beginner such as myself.

    Im interested in Noise Reduction and how you achieve this. I have been shooting at ISO800, F2.8, 10mm, D7000 and find the noise high. If I go to 3200 Id be worried about the noise level being unacceptable.
    Do you do your noise reduction in camera (LENR), ACR, Lightroon, P’Shop or 3rd party?
    Love to hear more.
    Kind regards

    1. Steven Christenson

      “I’d be worried about the noise being unacceptable”

      1. Have you tried it? The D7000 is a decent performing camera.
      2. When one “pixel peeps” there are often many things that you’ll notice – things you may never notice in the final image.
      3. My third comment, is I’ve been working for many months now on a “Down with the Noise” webinar. There is no one-size fits all answer, but I can say pretty confidently that “LENR” is the least effective method.

      In response to your email to me:

      The part of the Milky Way that you’re photographing is not the richest part. The exif data says 9:30 PM. While your photo shows part of the Milky Way (indeed EVERY star you see is part of the Milky Way), it is not the part in which the densest “clouds” are. See here:

      Other problems with your exposure: it’s IS0 800, and f/3.5 – you really do need to get up to ISO 2000 and f/2.8 – or better.
      For exactly what it looks like “out of the camera” see:

      By the way, I’ll be talking a bit about processing Milky Way Images in my next Webinar.

  8. Graham Wiffen

    Thanks Steven,
    I have signed up for the Reduce Noise webinar, whenever you run that and will look at the other one too.
    Will certainly try out the high ISO settings and see what I get. By the way, what are ‘pixel peeps’?

    1. Steven Christenson

      A pixel peeper is one who zooms in to nearly 100% (or more) and judges the overall quality of an image at that resolution. It’s like judging a forest by a single leaf. Pixel peeping is useful when gauging the effects of certain actions like sharpening and noise reduction; but it is a bad way to judge an image.

      The next worse way to judge an image is by “chimping”. Hunching over your LCD and admiring the shot. The LCD display is too small to show the level of detail (including noise, sharpness) or the quality of exposure. It’s a helpful guide, but the histogram is usually better – if you know how to read one. Fortunately an article on histograms is in the works – I haven’t found any on the web that adequately address night images.

  9. Graham Wiffen

    Hi Steven,

    Interesting, took a series of trial images last night at ISO3200 F3.5 (which is the widest it will go) and the Noise level wasn’t as bad as I had imagined. Not too much different from ISO800.
    Also played with different Noise Reduction approaches and so far, find the NR in ACR to be the best, but looking forward to you NR Webinar.


What do you think about this?

This site uses Akismet to reduce spam. Learn how your comment data is processed.