Wishlist: varying psf deconvolution coma correction

Firstly, just to say thank you for such awesome software. I’ve been using siril for a while. I recently started a PixInsight trial period and have been A/Bing some processing with PI and siril - results, at least for my setup, are actually in favour of siril except in a few cases. Certainly sufficiently so that I think I’ll be saving myself the cost of a permanent PI licence :slight_smile:

Here’s my wishlist item (and it’s something that PI doesn’t do either). When imaging with a Newtonian we have to deal with coma. Mostly it’s corrected by my coma corrector but if collimation is a whisker out there can be some residual aberration resulting in elongated stars near the corners.

I came across this tantalising web page outlining an approach to deal with it using deconvolution: Correcting Coma/Field-Curvature Type Aberrations With Deconvolution

Essentially the process is to measure the psf length, width and angle of a selection of stars at various points throughout the image and then use these measurements to identify the central point of the aberration and parametrise the psf as a function of radial distance from that central point. A deconvolution process is then carried out, but at each point in the image the psf used for deconvolution varies according to the measured parameters.

It sounds like a neat idea and the example images on the website certainly show that it delivers some improvement. Unfortunately the only implementation of it is for ancient versions of AstroArt.

It would be very nice if Siril could implement a process like this, as far as I am aware it would be unique among current astro processing software.

Not sure if this is what you have in mind https://www.startools.org/modules/sv-decon
Che

Ooh yes, that looks like what I was describing. I take it back about there being no current solutions then XD And pretty fairly priced too, I’ll have to look into getting that. Thanks for the tip.

Still, no harm leaving it on the siril wishlist too…

I came across this tantalising web page outlining an approach to deal with it using deconvolution: Correcting Coma/Field-Curvature Type Aberrations With Deconvolution

I found that post as well and contacted Professor Vanderbei to ask how he would feel about open sourcing his code behind that example - my intent originally being to put it into a GIMP plugin. He was very positive and helpful, providing the source code and advice. In the end the math was beyond me and I gave up, but Siril would be a good home for this functionality if someone more capable than myself can integrate it.

Here’s the source code that Professor Vanderbei sent me.
pideconv.cpp.zip (7.3 KB)

Wow, that would be a very good add indeed.

Thanks. I’m working on a simpler siril filter at the moment but if I can do that successfully I might have a go at this one next.

What type of filter?

The generalized hyperbolic transform stretch functions proposed by David Payne on the astrobin forums a while back, plus their inverses. I’ve just submitted the merge request.

1 Like

Those are great.

How will this work in Siril?

You can have a look at the current state of things in my repository on gitlab - Adrian Knagg-Baugh / Siril · GitLab
The master branch (which is what I’ve submitted for merging to the main siril repository so far) has a simple sliders interface, and the ght-experimental branch has a very rough first go at an interface with a histogram preview. Nothing’s decided yet about exactly what it’ll look like when it gets an official release in siril v1.2, but you can try out both versions using the gitlab repository. There are automated builds for win64 and linux appimage available in the CI/CD / Jobs menu (experimental branch ones on the front page currently, master branch is currently on page 2) or you can clone the repository and build from source if you prefer.