A Free Tool for Lucky Imaging

While I don't spend much time on image processing in this blog, I'm going to make an exception this time around. I've found a great new program for my tool chest when it comes to lunar and solar image processing, and I think you might like to try this powerful, easy-to-use tool for yourself.

We've already covered how lucky imaging involves recording videos of the Sun, Moon, or planet, and using software such as RegiStax or AutoStakkert to find the sharpest frames within that video. The software aligns and stacks those images together. Then, in the final step, the image is sharpened with wavelets. (Sharpening must be undertaken with care, as the process has an annoying tendency to turn noise into features that don't actually exist!)

For a while now, I've been using AutoStakkert to process the video files I captured with SharpCap. Then I switch to RegiStax for wavelet sharpening and use Photoshop for any final touches. I've found that using wavelets is a black art. Mastering the sliders in Registax takes practice; you need to experiment to get a feel for how they function. To many people I know, they are still a mystery.

Years ago, when I was on a team writing software for the refractive eye surgery market, I learned about using convolutions and deconvolution to simulate optical defects and corrections in a patient's cornea, and I've always thought these ideas could make a better approach to sharpen a blurry lunar image. But I never had the time to delve into the idea myself.

European software developer and amateur astronomer Filip Szczerek seems to have had the same insight. He has released an open-source software package (as in free, including with source code) called Image Post-Processor (ImPPG).

ImPPG applies Lucy-Richardson deconvolution (that's fancy math for 2D signals) along with some other processing to sharpen and otherwise improve images. While I can't specifically recommend it for planetary processing, it is effective and easy to use on high-resolution images of the Sun and Moon. I find it to be far easier and more intuitive than using wavelets in fact.

One of the highlights of this software is that it employs your computer graphics processing card (GPU) to speed up the process. So the same graphics technology that's used for epic orc battles in computer games can also speedily sharpen your blurry lunar and solar images.

The interactive nature of this tool makes it easy to experiment with its very few sliders and quickly get a feel for how best to sharpen your images. The default settings already produce an impressive result. But just as when you are using wavelets, resist the urge to oversharpen, and be on the watch for artifacts in noisier images. Sharp and smooth is the gold-standard to aim for.

Another thing I love about this software is that its curves tool is very easy to use, and allows you to adjust the contrast of your sharpened image as you go.

There's a great tutorial online here, but I'll summarize by saying the easiest workflow is to adjust the Sigma (the Lucy-Richardson part) first. Once you have a satisfactorily sharpened image, use the lower controls to just slightly apply an unsharp mask. You'll find that, most of the time, the sigma for the deconvolution and the unsharp masking is the same value.

I find this so much easier and simpler to use than wavelets and am very happy with my final lunar/solar images. The "deringing" option is only needed when you have solid white or solid black edges. But even with this enabled, I don't find planets process well this way. The finer-grained control of wavelets does win in those cases, as it helps squeeze out the ultimate details.

There is one potential showstopper for some imagers, and that is that this tool will only work on monochrome images. This is not a problem for most solar and lunar images, though. Those imaging with color cameras can still bring color images into ImPPG, they're just converted to monochrome. Or, you can split your color channels in Photoshop and process each one in ImPGG individually.

The source code to this tool is also available for free, and prebuilt binaries (the executable) are available for Windows and Linux. Check it out!