Registration strangeness


I’m currently registering and stacking nearly 800 multisession subs from 2 different (but close) image scales and have come across a weird registration anomaly.

Some of my subs have not registered correctly and look like the example shown.

It is not the case of only one image scale registering, as most from the same camera register fine with the other camera. The pre-processed garbled subs even look the same as the rest of the subs.

Any idea of the cause?



Hi, I haven’t seen this recently, if your input images look normal this is very strange… Does it happen always with the same images? Could you share the reference image and a failing image? Thank you

Registration failed and a bad transformation matrix was computed.
Using affine transformation instead of homography could help to exclude frame.

Hi Vincent,

Here is a link to 2 files - one good, the other bad.

It has happened on occasion before, but I usually end up just excluding or deleting the bad files.

Hi Ciryl,

Sorry, I’ve no idea how to “affine transformation instead of homography”.

What is this in layman’s terms, please?



I tried your images, they are very different in terms of level, stars still seem to be detected correctly (more than 1400), but no registration solution is found. Are you sure they are of the same sky area?

What Cyril was talking about is an option in the registration algorithm, available in the GUI or as an option of scripts (-transf, see the command reference)

About your pictures… The images are so different, not even the same size.
Then you reference image is not correctly demosaiced.

So, your input images are not good. Difficult to do something.

This is tipically a pattern when the chosen Bayer pattern was wrong:

That is weird - there is no option for that in my version of Siril 0.99.11?

0.99.11, i.e dev version, is updated almost everyday. Your version is too old (for a dev version)

But your main problem is not that.

How do you update my version when update function doesn’t work in my version?

Why is it not correctly debayering, when the debayer settings are at default RCD?

You are probably on Linux, with an outdated version of libraw that does not take into account your DSLR.
But this is just hypothesis.

Nope, windows 10.

Hypothesis remains. Need to update msys2 dep packages

Will this fix the debayer issue?