Sharpening the moon
How to edit your moon photography
Photographing the moon is easy, as long as you have a good tele lens, a stable tripod and a clear night sky. But making a sharp photo of the moon is surprisingly difficult.
Light pollution and the natural turbulence of the atmosphere interfere, and while your eyes (or rather you brain) compensate for this, a static photo of the night sky is often blurrier than you would expect. As Heraklit wrote: Pantha rhei (everything is in motion), especially celestial bodies. The rotation of the earth results in light trails, as soon as you expose for more than a few seconds, even though the stars are far away and their own movement isn’t noticeable for us. The moon however is literally speeding across the sky in its orbit around earth and even at exposures well under a second this motion adds blur to a photo.
The following picture was taken on a clear, cold night and focused as precisely as was possible with live view magnification and focus peaking (which highlights sharp edges):
Stacking bits
The classic astrophotography procedure to get sharp images with lots of details consists of making lots of images and stacking them in a specialized software, removing noise and recovering details in the process (and revealing extremely dark objects like nebula, when you are not photographing a lightbulb in the sky like the moon). Fortunately Affinity Photo, an excellent and cheap alternative to Photoshop, has a dedicated astrophotography mode that takes a lot of the pain out of using a specialized software. The result of this stacking process is astounding:
Suddenly a heap of blurry pictures becomes one sharp photo of the moon. The only problem is that handling 50 RAW files (about 40 MB each) needs lots of memory and processing power. Even my iMac Pro took a few minutes to merge all this data. But the result speaks for itself. Had I used the K-1s Pixel Shift mode, the resulting image could have been even sharper, but it would have meant stacking 50 RAW files of about 150–200 MB each. As you can see, the process, even if it’s worth it, is quite challenging.
Surely there must be a better way? I’m glad you asked!
AI is the solution (or is it?)
Modern image processing software is astonishingly good at improving your photos. This is also true for sharpening blurry pictures, thanks to recent advances in artificial intelligence (long for AI). My favorite tool to sharpen and denoise images is DxO PhotoLab. Its DeepPRIME technology gets rid of any noise and in the process sharpens your image by trying to restore blurry parts using an AI algorithm. Additionally its classic (non-AI) sharpening tools work extremely well — better than any competing product in my opinion. Here’s the result I got when I processed the original (blurry) image with DxO:
If you compare it with the original, the improvement is really remarkable. Here is a direct comparison of DxOs result with the stacked version:
Take a moment to admire how well DxO PhotoLab reconstructs missing details from a single, unsharp picture. It’s not quite as good as the stacked one, but it’s a close call. The stacked image ist still a bit sharper and shows details that are missing in the version sharpened by AI, especially in larger, darker areas, which tend to be smoothed out by the software.
Even more intelligent AI
While DxO uses AI to remove noise and add detail, the above image still needed quite a lot of manual adjustments, especially in setting the sharpening parameters. This processing takes a certain amount of time and experience and is by no means easy to achieve. To quote the Beach Boys: Wouldn’t it be nice… if AI would do the whole job automatically, in a one-click process?
That’s exactly what Topaz Sharpen AI promises to do. Sharpen AI uses machine learning algorithms to recognize elements of an image and restore them. By doing so it sometimes invents details that weren’t really there. This can work wonders, for example in furs, feathers or for human faces. And it can miserably fail when it tries to sharpen something that in reality wasn’t sharp. Additionally, it doesn’t seem to be optimized for astrophotography and night skies can turn out noisier than they were in the first place . But thankfully for moon photography (where the night sky is blown out by the bright moon anyway) the app works really well. A clever comparison view in the app shows the results of three different processing algorithms side by side and you just select the one that works best (which was “motion blur” with this particular photo, but may be “too soft” or “out of focus” for others). You can fine-tune the results if you wish, but the “auto” setting usually is already very good. So it really is a one-click sharpening tool and the result is almost as good as the one from DxO:
On the one hand, Sharpen AI added a slight yellow color cast (which can easily be removed), on the other hand — while the image may seem even sharper than the result from DxO — it obviously added details out of nowhere (e.g. in the area around the Picard crater on the top of the photo). And by doing so it actually made some existing smaller craters disappear. So while the result is an attractive and highly detailed image of the moon, it’s not exactly a scientifically accurate representation of the real thing.
Conclusion
Is AI a viable alternative to the time-consuming and complicated process of stacking dozens of images? Can AI restore a blurry picture to all its unblurred glory? Yes, to a certain extent it can do this. But only if you are willing to sacrifice some of the quality you would get from the stacking process. And more importantly: you shouldn’t do it if you want to keep your integrity as a photographer who documents reality, as AI is not only artificial in itself, but the results of its work are unreal too.
If you have only one picture you want to improve, go for it! If you have the time and patience it needs to get the best possible result, it’s worth taking the long road.
