PIPP, AutoStakkert and RegiStax in One Window

If you have processed a planet before, you know the sequence. Open the capture in PIPP to center and crop it and write a new SER. Open that in AutoStakkert to grade the frames, set alignment points, stack the best fraction and write a TIFF. Open the TIFF in RegiStax and pull the wavelet sliders until the belts appear. Three applications, two intermediate files, and settings in each that you learned once and now copy from a forum post.

None of the three is bad software. Together they are why planetary imaging feels like a craft with an apprenticeship. This post walks through the chain step by step and shows where each step lives in Akastroid, and also what the originals still do that Akastroid does not.

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.

Step one: PIPP (center, crop, discard)

PIPP’s job is preparation. It finds the planet in every frame, moves it to the center, crops the frame to a box around it so the stacker is not aligning on empty sky, throws out frames where the planet clipped the edge or the exposure jumped, and writes a clean SER for the next tool.

In Akastroid this happens when a capture is opened, and there is no file at the end of it. Every frame is located (the disc’s centroid, measured to a fraction of a pixel), and its brightness inside the disc is recorded alongside its sharpness. A passing cloud or a gain change shows up as a brightness dip and the frame is ranked accordingly. Frames where the planet ran off the sensor are not usable and are marked as such. Cropping to the planet is a button on the result, and it is optional because some people want the moons in the frame.

What you lose: PIPP’s output options. It can write an AVI for a different program, split a long capture into chunks, or produce a centered animation. If PIPP was a hub in your workflow rather than one step in it, that hub is not here.

Step two: AutoStakkert (grade, align, stack)

This is the step that makes the image. AutoStakkert measures how sharp every frame is, sorts them, lets you choose what fraction to keep, places alignment points across the disc, and then stacks each point’s patch from the frames that were sharpest there rather than sharpest overall. That idea, local quality ranking, is why it displaced everything before it.

Akastroid does the same three things, with two differences worth knowing about.

Grading measures each frame’s fine-scale structure relative to its own noise, rather than raw edge energy. Raw edge energy rewards noise: a noisy frame has plenty of high-frequency content and none of it is Jupiter. The ranking is shown as a frame browser you can scrub through from best to worst, which is the quickest way to see what the atmosphere did to your evening.

The keep fraction is read off the capture, not typed in. The conventional “best 25%” is right for some captures and wrong for others, and the data tells you which. On a Jupiter capture in steady air, the top quarter of frames all scored 75 or better against the best frame’s 100, and stacks at 25%, 12% and 6% were indistinguishable. On a Saturn capture in rough air the top quarter reached down to 48, with the median frame at 40, and 12% gave measurably sharper belts than 25% for a small cost in noise. Akastroid keeps the frames that score close to the best and never fewer than half, because cutting to the conventional quarter measured worse on both sharpness and noise, and shows you the number so you can override it. (More on how that was measured.)

Alignment points are placed automatically on the parts of the disc that have structure to align on, such as a limb or a belt edge, and each point ranks the frames locally. AutoStakkert keeps a fixed fraction at each point. Akastroid weights continuously instead: a frame that was sharp at a point contributes fully there, a frame that was soft contributes in proportion, and there is no cliff where frame 499 counts and frame 500 does not.

What you lose: manual alignment-point placement, and scale. AutoStakkert lets you put points exactly where you want them, and on a very long capture (ten thousand frames of a large, detailed disc) its local ranking has more to choose from than Akastroid’s does. Measured on a modest capture the difference is a percent or so. On a long, deep one AutoStakkert is still the reference.

Step three: RegiStax (wavelets)

RegiStax’s wavelet sharpening is why the technique produces images that look the way they do. The stack is decomposed into layers by spatial scale, each layer is amplified by a slider, and the image is rebuilt. Detail at the scale of a belt gets lifted, and noise at the scale of a pixel does not have to be.

Akastroid’s sharpening panel is the same tool with the same six scales, from Finest to Coarsest, each with its own noise suppression. The starting values are not a preset; they are tuned from the stack. A deep stack in good seeing gets its finest layer lifted, because at that depth the finest scale holds detail rather than noise. A thin or soft stack does not. The sliders re-render immediately without redoing the stack.

Two things RegiStax users reach for are also here. RGB align measures the offset between the color channels and corrects atmospheric dispersion (the red fringe on one limb and blue on the other that a low planet always has) to a fraction of a pixel. Restoration is a deconvolution with the blur measured from the planet’s own limb. It is off by default, because deconvolution invents plausible detail with the same confidence it recovers real detail, and a planet has no star in the frame to check the answer against. Raise it slowly and compare.

What you lose: RegiStax’s linked wavelets, its histogram and RGB balance tools, and its per-layer preview. If you have a RegiStax recipe you trust for a particular target, the sliders here will get you close, but they are not a port of that recipe.

What none of this replaces

WinJUPOS de-rotation. Jupiter rotates fast enough that a capture longer than about ninety seconds smears its own surface. WinJUPOS compensates, so you can stack across ten minutes instead of two. Akastroid has nothing equivalent, and for serious Jupiter work WinJUPOS remains the tool. It is free.

Batch depth. AutoStakkert will work through a folder of fifty captures overnight with one settings file. Akastroid has no batch mode; sessions are processed one at a time.

Summary

The three-application chain exists because each step was solved separately, by different people, at different times. There is no technical reason for the seams between them, and the intermediate files are a historical leftover. Akastroid removes the seams and takes the decisions that always depended on the data (the keep fraction, the alignment points, the starting wavelet values) from the data.

If you are learning, that removes most of the apprenticeship. If you already have a workflow that produces images you are happy with, the reasons to switch are convenience and the frame browser, and the reasons not to are manual alignment points and de-rotation. Both are real, so weigh them against how you work.

Compared directly: Akastroid vs AutoStakkert and Akastroid vs RegiStax, measured on the same captures. More on the planetary side: planetary stacking software.