Help with manual registration

I’m trying Manual Registration of three light frames. (it’s a test)
The first image is the reference image, so I click on the second, set both first and second preview and then adjust the XY translation. As soon as I start adjusting, the third image (the last one) is loaded.
When I finish adjusting and click on the second image, it is not adjusted.
In summary, the issue I’m facing is that apparently Siril doesn’t allow me to adjust translation of images different from the last one.
I’m assuming that Manual Registration needs to be done for each image on the sequence.
I read the manual but couldn’t find the answer.
DISCLAIMER: I’m not an astrophotographer and the images have few stars and are surrounded by the scene foreground either at bottom and on both sides, which seems to prevent automatic registration (it gives false positives probably because of the foreground).

Which version do you use? 0.9.12 has a big issue with it

Ah, ok, mine is 0.9.12.
Would downgrading to 0.9.11 be ok?
I should have to build it from source, correct? Or is there some appimage lying around?

ok, please disregard the question about appimages, I found them on the site for previous versions.
I’m running 0.9.11 appimage and this time the Manual Registration behavior is as expected.
After manual registering four images, Siril seems to have taken into account pixels from the foreground as stars. I came to that conclusion by clicking on each image after registration, and see that the stars move (they shouldn’t, right?)
So I’m trying to mask out the foreground by selecting Match stars in selection e dragging an area in the sky. However, it warns me that the area is too big. When I reduce the selection to an acceptable size, it has very few stars.
Is there any other setting I can tweak to allow Siril to register a polluted sky while masking out the foreground?

Why don’t you try the one star alignment?

I tried, but then almost all (few…) stars vanished, since there’s some kind of rotation going on. Only a region around the star I’ve chosen were visible.

If you have field rotation, manual alignment won’t work szdly:(

ok, @lock042, thanks for your attention.

I guess I’ll have to look for some (much more) darker sky.

@lock042 I’m thinking, more as a feature request, does it make sense that manual registering uses the same algorithm as global aligment, except for the star lookup part of the algorithm?
This way, Siril would fit also poor skies like mine + field rotation - of course, assuming that I would profit with some improvement, like revealing some more stars (10% more?) and reducing some noise.

We could imagine a 3 star manual registration algorithm to handle rotation. On each frame we peak 3 same stars.
Why not.

1 Like

Great! Thanks, @lock042.

I opened a feature request in Gitlab: