Anybody interested in processing an atypical M51 stack? - Forums

Topic starter

Jul 23, 2021 · Edited: Jul 23, 2021

I'm getting my kicks by processing my images using my own Python programs. Not talking about processing skills, this means that the end results may not be as good as they could be if processed with better tools. I am thus very interested in seeing what can be obtained from one of my stacks with, e.g., PixInsight or Astropixel Processor (or any other state of the art S/W). I could try that myself but I don't know either one and would not know how to best use them. This stack:

https://www.dropbox.com/s/u7wazbbbxg3hl5m/M51_astrobin.fts?dl=0

is the source for this M51:

I call it atypical because it was obtained with a Samyang 135. So not the best image of M51, but I'm playing with getting the most out of a little lens.

My version of the stack was obtained by wavelet regularized Richardson-Lucy deconvolution followed by luminance stretch, saturation and two passes of unsharp mask. For the deconvolution, I did the RGB channels separately (with PSF estimated from stars) and ran 50 iterations on the red channel and 20 on the other two. My deconvolution is a bit heavy and I do have some ringing around the stars. I'm sure there is room for improvement.

So if anybody has time to spare (hum), I'm very curious to see what is possible. Feel free to use the fits anyway you like!

CS

Frédéric

Helpful Insightful Respectful Engaging Supportive

David Nozadze

Jul 23, 2021

How interesting! Downloading now. Will give it a try

Frédéric Auchère

Topic starter

Jul 23, 2021 · Edited: Jul 23, 2021

David Nozadze:
How interesting! Downloading now. Will give it a try

Thanks for taking some of your time! I've added a few more details on my processing in my original post.

Well Written Respectful

David Nozadze

Jul 23, 2021

It looks like you developed an excellent process! I could not get anywere close to the same quality, with several manual or automated methods. So, congratulations!

Claudio Tenreiro

Jul 23, 2021 · Edited: Jul 24, 2021

Frédéric Auchère:
I am thus very interested in seeing what can be obtained from one of my stacks with, e.g., PixInsight

Hello, I will send you soon one possible result using PxI. I believe that the main difference will be in the stars. Just give me an hour or so.
Cheers and CS!

Ok, Done in a fast processing mode, without re sharpening using any aggressive process to make it more sharp, so it is possible still to do some colour enhancement and making it more sharp, but that takes more time to achieve the desired level of contrast and so on.

Data From Frederic

After you see it, let me know to delete it (it is not visible right now).

Claudio

dzambon

Jul 24, 2021

Frédéric Auchère:
...processing my images using my own Python programs.

Curious to see what your code does

do you have it in any public repository?

Mark Germani

Jul 24, 2021

Hi Frédéric:

Even though I don't completely understand the complex deconvolution you've performed, I'd love to have a go at stacking this in Astro Pixel Processor with some aggressive drizzle settings, since you've got at least 4193 subs (wow!) My own attempt at M51 is similar in that I'm working with a shorter focal length, and 45s is near enough to lucky imaging. I gathered a tenth of the subs you did, so I don't have anywhere near the detail you've achieved, so I'd love to see what I can get from yours!

At 135mm on the ASI178, you're massively undersampled, and I see you weren't guiding so you weren't manually dithering, but I often find that I get enough drift over time without guiding so it's sort of dithered anyway ;) So all of this suggests that drizzling could potentially bring out a lot of extra detail.

If you're interested, there's one small hurdle - I'd need all your light frames, and your calibration frames, and that's a lot of data to upload to Dropbox. If you are able, somehow, to share the data with me, I'd love to give it a go! Perhaps we could upload/download the subs in a few large chunks so as not to exceed Dropbox limitations?

Cheers,
Mark

Helpful Insightful Respectful Engaging Supportive

Frédéric Auchère

Topic starter

Jul 24, 2021

Claudio Tenreiro:
Ok, Done in a fast processing mode, without re sharpening using any aggressive process to make it more sharp, so it is possible still to do some colour enhancement and making it more sharp, but that takes more time to achieve the desired level of contrast and so on.

Data From Frederic

After you see it, let me know to delete it (it is not visible right now).

Thanks very much Claudio for taking the time to do this! I've checked the result, you can delete it if you wish. I like the tint you get. If put my deconvolved version on dropbox in case you try some yourself.

Frédéric

Frédéric Auchère

Topic starter

Jul 24, 2021

Frédéric Auchère:
...processing my images using my own Python programs.

Curious to see what your code does do you have it in any public repository?

Sorry, I don't have it in a public repository, only the wavelet part. The main processing step is deconvolution, and I put the core code here. From what I understand from the docs, it is roughly similar to what is done with wavelets regularised deconvolution in PI (e.g. here).

Frédéric

Frédéric Auchère

Topic starter

Jul 24, 2021

Mark Germani:
If you're interested, there's one small hurdle - I'd need all your light frames, and your calibration frames, and that's a lot of data to upload to Dropbox. If you are able, somehow, to share the data with me, I'd love to give it a go! Perhaps we could upload/download the subs in a few large chunks so as not to exceed Dropbox limitations?

Hi Mark,

Thanks for you interest. Indeed we're trying to do something similar on M51

In the raw stack, I'm getting stars about 1.4 pixel wide, so undersampled but not that much. In these conditions, drizzle is pretty effctive. The stack is heavily drizzled (x4 resolution with 50% droplet size), with dithering at every frame.
I'd love to share the subs. I'll get back to you once I figured out where to put them (about 60 GB).

Frédéric

Helpful Respectful Concise Supportive

Mark Germani

Jul 24, 2021 · Edited: Jul 24, 2021

Frédéric Auchère:
The stack is heavily drizzled

Ok, I wasn’t sure if you had already drizzled (I see it now in your notes). If that’s the case, I probably can’t tease much more out of it than you did. Probably not worth trying to transfer 60gb of subs.

Fantastic work, thanks for sharing!

CS,
Mark

Frédéric Auchère

Topic starter

Jul 24, 2021

Mark Germani:
Probably not worth trying to transfer 60gb of subs.

OK let me know. I think it is possible via "dropbox transfer".

Die Launische Diva

Jul 24, 2021

@Frédéric Auchère, the stack is very good and clean. I did a preliminary processing in PI and I am impressed. More to come later!

Die Launische Diva

Jul 24, 2021 · Edited: Jul 24, 2021

And here is my result:

No noise reduction applied!

Very nice dataset, for any problems please blame my PI skills and not the data. Someone who writes his own code for his own data probably will become a successful astrophotographer. I tried to respect your data, Frédéric (well OK, I may went too far with Deconvolution)!

Frédéric Auchère

Topic starter

Jul 24, 2021 · Edited: Jul 25, 2021

Die Launische Diva:
No noise reduction applied!
Very nice dataset, for any problems please blame my PI skills and not the data. Someone who writes his own code for his own data probably will become a successful astrophotographer. I tried to respect your data, Frédéric (well OK, I may went too far with Deconvolution)!

Thanks very much! That is really good, wow! I'm learning a lot from this

As you can see, I have myself problems with deconvolution over a larger FOV:

Your deconvolution seems better controlled. May I ask you a few details about it? My bright stars are ugly but that's from saturation, I suppose I could fix that. In the outer FOV however I tend to get nasty artefacts, maybe from a bad PSF model. You wrote in your description that you get the PSF from stars picked in the FOV. Is it then an average of the stars? A model fitted to that? You also wrote that you did not use noise reduction but I suppose there is some regularization applied during deconvolution? I may also be pushing the deconvolution too much. How many iterations did you use?

I really appreciate the feedback

Frédéric

Related discussions

Optimal data flow from remote observatory?

Attached to one of the latest great images from @John Hayes, he described a recent improvement in his dataflow. After sharing some thoughts back and forth, we thought this might be a good topic to bring to the forum for a bit of a wider discussion an...

Jul 14, 2025

Both posts discuss sharing and improving astrophotography image processing workflows and techniques within the astronomy community.

Lessons learned & tips from building an observatory (remote or not)

Hi all, I wanted to start a topic specifically for those of us who either have an observatory or are thinking of building one be that remote robotic or not. Whether you're in the planning stages or already running a fully automated system, I beli...

Jun 9, 2025

Both posts involve astronomy enthusiasts seeking to learn from and collaborate with the community to improve their astrophotography capabilities.