How to register with large atmospheric distortion? [Deep Sky] Processing techniques · HR_Maurer · ... · 5 · 257 · 0

HR_Maurer 2.86
...
· 
·  Share link
After capturing a lot of imagery with comet C/2023 A3, i'm struggling to stack those images because of imperfect registration. This comet has been standing very low in the east during dawn.
The subs i want to stack were shot with a Sony a7iii and a 135 mm lens, mounted on a star tracker. Nearly all include the horizon, which is ~1.5° below the point of observation. Looking a loong way through atmosphere. There is a large area of ~undistorted star field in the images, which can be registered. Towards horizon, distortion increases and becomes very large. Since the images are tracked and stars were rising, the distorted area is different in different subs. And on top of that, dawn has begun and there are large and changing gradients in the sky.

I've been trying to tweak Image Registration parameters, and also tried repeated registering, but none of that worked so far. I think the registration model is important, as well as the method of interpolation. But im not 100% sure if the stars in the brighter image areas have been detected at all.
Ive been trying to create an undistorted artificial reference. But so far ive been failing to plate solve one of the subs. So, at the moment i ran out of ideas.
Edited ...
Like
andreatax 9.89
...
· 
·  Share link
Manual registration.
Like
ScottBadger 7.63
...
· 
·  Share link
In Pixinsight, DynamicAlignment.

Cheers,
Scott
Like
HR_Maurer 2.86
Topic starter
...
· 
·  Share link
Okay 👌 
I did this before. This is the last option. I'd rather write my own registration code in Matlab.
Like
Tromat 0.00
...
· 
·  Share link
"Thin plates splines" instead of standard "projective transformation" in Pixinsight should be able to manage this.
Like
HR_Maurer 2.86
Topic starter
...
· 
·  Share link
Okay, so here's what i did for proper registration:

1. calibrate (of course)
2. remove background gradient
3. eventually remove median
4. Register, using Thin Plate Splines, and star matching: Quadrilaterals. I used 8 detection scales under star detection for my specific images

worked.
Edited ...
Like
 
Register or login to create to post a reply.