How to Process Astrophotos Without Learning PixInsight

“Learn PixInsight” is the standard answer to “how do I process my astrophotos”, and it is not bad advice. It does come with a cost that rarely gets mentioned: several months before you have a photograph.

For some people in this hobby that cost is clearly worth paying. A much larger group just wants a good image of the Orion Nebula from the frames sitting on their drive, and three months of learning a signal processing environment is a lot to ask of them. This is written for that second group.

Our recommendation

For most people the answer is Akastroid: one app for deep sky, planets, the Moon and the Sun, on Mac, Windows and Linux. You drop in your frames and it grades, calibrates, stacks, removes the gradient, calibrates the colour and sharpens, with no settings to choose. The usual route needs four to six separate programs for the same result.

On the same Orion stack, Akastroid’s automatic finish measured a signal-to-noise of 402 against 87 for Siril’s standard workflow, with a sky gradient of 2.6% against 21.6% (the full comparison). It is free to try on your own data.

Choose PixInsight if you want to control every step by hand and will spend months learning it, or Siril if you want free software and are happy to learn the steps. The rest of this guide compares them in detail.

What processing is

Strip the software away and the job is eight operations, in this order:

1. Calibrate. Subtract what the camera added. Darks remove sensor noise patterns, flats remove dust shadows and vignetting, bias removes the read offset.
2. Grade. Find the frames that are ruined: cloud, satellites, lost focus, a tracking slip, someone’s headlights.
3. Register. Align every frame to the same stars, to a fraction of a pixel.
4. Integrate. Combine them, rejecting outliers, so the noise falls and the signal does not.
5. Remove the gradient. The sky is brighter on one side than the other, and that has to come out without taking the nebula with it.
6. Calibrate the color. Decide what white is, ideally by matching stars in your frame against a catalog of stars with known colors.
7. Stretch. The data is linear and the faint parts are invisible. This is where an image appears, and where most beginner images are lost.
8. Sharpen and denoise. Carefully, and less than you want to.

Every program does these. What differs is who supplies the numbers.

What PixInsight is for

Before going further, the case for PixInsight deserves to be stated properly.

Its real advantage is masking. Range masks, luminance masks, star masks and PixelMath let you apply a correction to part of an image: stretch the outer regions without blowing the core, reduce noise in the background only, sharpen the structures that can take it and leave the ones that cannot. That is what takes an image from processed to finished, and nothing automatic replaces it.

If you look at award-winning astrophotographs and want to make those, you will end up in PixInsight eventually, and you should. What follows is about getting from a folder of frames to a real photograph without PixInsight being step one.

Route one: Siril, with its scripts doing the middle

Siril is free, open source, and runs on Windows, macOS and Linux. Its one-shot color preprocessing scripts will calibrate, register and stack a session without you specifying much, which covers steps 1 to 4.

You then do 5 to 8 by hand in Siril: background extraction, photometric color calibration, a histogram or generalised hyperbolic stretch, and whatever denoise you want. That is a real skill, but a much smaller one than PixInsight, and Siril’s documentation is written to be read by people learning it.

Cost: nothing. What it asks: that you learn six or seven operations and roughly what their parameters do. Where it stops: the scripts are fixed. They apply the same settings to every session because they never look at the data, so a wide field and a tight crop get the same black point, and a night where the seeing collapsed at midnight gets all its frames stacked anyway.

Route two: GraXpert and Photoshop for the finishing

If you already know Photoshop, a workable route is DeepSkyStacker or Siril for steps 1 to 4, GraXpert for step 5, and Photoshop for 6 to 8 with curves and levels.

Cost: whatever Photoshop costs you. What it asks: existing Photoshop fluency, plus enough astro knowledge to know that your normal photo-editing instincts are wrong here. The background is not supposed to be black, and a curve that looks right on a daylight image will destroy faint signal.

Route three: have the measurements made for you

Those eight steps take months to learn, but not because any one of them is conceptually hard. Each has parameters, and knowing what to set them to is pattern recognition built up over dozens of sessions.

Almost none of those parameters are matters of taste, though. Which rejection method suits a stack depends on frame count and variance. How hard to deconvolve depends on measured star width and signal-to-noise. Where the black point belongs depends on what fraction of the frame is sky. These are measurements, and a program can take them.

That is what Akastroid does. We make it, so weigh this accordingly. It reads the folder, measures the session, runs all eight steps from those measurements, and lists every decision with the number behind it so you can overrule any of them. A hundred-frame session takes under a minute. The free tier gives you three processing runs of the full pipeline and one export before it asks for a licence.

Cost: free with every feature and one export, then $79 once. What it asks: a folder. Where it stops: the adjustments are global. No masks, no selective editing, which is the same limit PixInsight exists to get past.

The combination that suits most people

This arrangement gets little attention and suits the most people, including plenty who already own PixInsight: use something automatic for steps 1 to 4, and finish wherever you like.

Calibration, grading, registration and integration are mechanical. There is a correct answer, it is determined by the data, and doing it by hand gains you nothing except having done it by hand. Akastroid’s Save the stacked image, unprocessed writes the integration before any finishing is applied, which is what PixInsight or Photoshop want as a starting point.

You then spend your time on steps 5 to 8, where the judgment is, in whatever tool you know. And you spend it on data you already know is worth the effort, because the automatic pass produced a finished version in three minutes and told you what was wrong with your night.

When you do need to learn the deep tool

Be honest with yourself about whether any of these apply:

• You want masks and selective corrections. Nothing automatic gives you these.
• You are producing images for competition or publication.
• You are doing narrowband combination with custom channel mapping and want control over every stage.
• You enjoy processing for its own sake. That is a perfectly good reason, and quite a few people are in the hobby for it.

If none of those describe you, the months are not obligatory. For a long time they were the only route anyone had.

Where to start

Take one session you have never managed to process. Run it through whichever route above fits your situation, and get a finished image out of it, however imperfect.

Then read what each step was doing. Looking at a result and working backwards is a quicker way to understand processing than learning the theory first and hoping an image appears at the end.