 ...
·
·
1
like
|
|---|
|
Hello and welcome I'd like to ask everyone who is using Siril and did Deconvolution processing on their light images or any kind of images. Does that work? Whenever I try to use that, Siril keeps saying Deconvolution failed and nothing to import. What does this mean? Is there something I am missing here, or what Do I do In this case? Cause I know this menu is responsible for the enhancing the quality. Is there a way to fix this, or forget about it cause it's useless? |
|
0.00
...
·
·
2
likes
|
|---|
| Every time I run it, it completes, but I never see any visible change in my image. Maybe my images are just awesome to begin with! 😄 |
0.00
...
·
·
5
likes
|
|---|
| Take a look at the YouTube channel Deep Space Astro, he has a great video on using deconvolution. It does show improvements, albeit minor. |
3.81
...
·
|
|---|
| If you can't see the change during processing just do an A-B comparison. In my experience, it's pretty subtle, which is what you want. |
2.11
...
·
|
|---|
| Do you need deconvolution for the complete image or are you primarily only trying to fix the stars? If stars only I often use Full Resynthesis. |
...
·
|
|---|
Do you need deconvolution for the complete image or are you primarily only trying to fix the stars? If stars only I often use Full Resynthesis. Kinda both but I mostly seeing people just want to make the starts look better. |
4.06
...
·
·
3
likes
|
|---|
|
Read through this: https://siril.readthedocs.io/en/latest/processing/deconvolution.html You should read through the rest of the documentation too, i have noticed that you post these questions regularly when basically all the answers can be found in the documentation. Just read it! |
1.91
...
·
·
4
likes
|
|---|
I was just playing around with this today. Try this:
|
|
0.90
...
·
·
3
likes
|
|---|
Szijártó Áron: I've never come across a situation yet where the deconvolution process completely fails. I can't find the exact error messages you mention anywhere in the source code, however if the second message is "No FFT wisdom to import" that's nothing to be concerned about, it's just a status message. It means that Siril hasn't yet saved any FFT "wisdom" (basically a cache of FFT planning data that helps the FFT library run more quickly) and therefore there isn't any to load in. The wisdom will accumulate over time for different image sizes you process. The default deconvolution settings are very subtle (with the benefit of hindsight, perhaps too subtle - maybe I'll change the defaults for the next major version). You can easily confirm if it's doing something by selecting the Multiplicative method instead of Gradient Descent in the "Algorithm Method" setting: with more than a few iterations you'll quickly see the ringing artefacts typical of over-deconvolved stars. Once you've confirmed it works by over-doing it, a good place to start is go back to the defaults and increase the gradient descent step size by a factor of 10. If that's still too subtle, increase it again, and so on. If none of that works I'd be very interested in a bug report, because I've never come across a failure like the one you describe at all. The best way to raise a bug report is to go to https://gitlab.com/free-astro/siril You'll need to make an account but once you've done that you can open an issue. Ideally please copy and paste the whole of the relevant log from the Console tab and describe what you did step by step so that I can try to recreate the issue. If you can post a link to the image you were working on, so much the better. |
|
0.90
...
·
·
1
like
|
|---|
Every time I run it, it completes, but I never see any visible change in my image. Maybe my images are just awesome to begin with! 😄 That's probably my fault for setting the defaults so conservatively. I didn't want the default to be prone to causing ringing artefacts and everyone thinking it was a rubbish implementation but perhaps I went too far the other way. Have a play with the gradient descent step size - try increasing it by a factor of 10 from the default, keep doing that till you see artefacts then back off a bit, then play with the number of iterations and see what works best for your typical data. |
|
0.00
...
·
|
|---|
| Thanks! Will check it out. I always just assumed I didn't have much need for it, but I've also never pushed the sliders before. Maybe I should. |
0.00
...
·
|
|---|
| Deconvolution has very little effect, it doesn't compare at all to deconvolution in PixInsight. I am overusing this feature in Siril. |
|
0.00
...
·
|
|---|
Szijártó Áron: |
0.00
...
·
|
|---|
| This youtube tutorial might help you. https://www.youtube.com/watch?v=dC5hyDeGhKk |
|
0.00
...
·
|
|---|
Deconvolution has very little effect, it doesn't compare at all to deconvolution in PixInsight. I am overusing this feature in Siril. I have found it easier than PI and have been happy with the results, and I find the more I use it the better it gets. |
0.00
...
·
|
|---|
Арон Сийярто (Arons.2001)Здравствуйте и добро пожаловать. Я хотел бы спросить всех, кто использует Siril и применял обработку Deconvolution к своим световым изображениям или любым другим изображениям. Это работает? Всякий раз, когда я пытаюсь это использовать, Siril постоянно говорит, что Deconvolution failed и ничего не импортируется. ***https://siril.readthedocs.io/en/latest/processing/deconvolution.html Maybe you should read the documentation?*** |
...
·
|
|---|
Adrian Knagg-Baugh:Szijártó Áron: Ah so this just only means that siril cant or couldnt found any data to save into the image. Could that be related to the resolution of the image im assuming. |
|
0.90
...
·
|
|---|
Keith Belzner: I'm a big fan of his channel, but it seems a slow and convoluted process so I don't bother. My images are rubbish though, so...... |
1.51
...
·
·
1
like
|
|---|
| I have had decent results from Siril Deconvolution, using PSF from stars. I sometimes save the PSF and apply it again with more iterations after star removal. |
0.00
...
·
·
1
like
|
|---|
| Siril Deconvolution works great at least for planetary imaging under good seeing conditions, with massive oversampling by 5x Barlow: I used AstroSurface for stacking with 32 bit *.fits result (AutoStakkert 4 with fits result also ok, but not as smooth as AstroSurface), and applied a large star PSF taken with same 5x Barlow, with kernel size 169x169 px covering the first two diffraction rings, then Split Bregman with high alpha values. See my latest images on Jupiter, Uranus and Mars, I don't use Registax nor other wavelet tools any longer, also no Photoshop etc.; just the deconvolution and FitsWorks. In my opinion, most important thing is to use 32 bit fits format for both the stacking result and the PSF in order ot avoid numeric artifacts. But did not yet try it on Deep Sky with native telescope resolution. |
0.00
...
·
·
2
likes
|
|---|
| tl;dr; i made a denoise/deconvolution-ish tool https://ubersmooth.com to help with this in my images (free download for windows, put in 0 and click the paypal button [no need for any accounts]). it is bare bones, command line interface, no graphical mouse stuff. background I used to use the Richardson-Lucy Deconvolution in siril, early in the linear workflow stage. I'd basically crank it up as much as possible before big rings around the stars, then back it off a little and call it good. It does sharpen everything up a bit, reduce star size a bit (not so good at fixing mis-shapen stars), but increases high frequency noise as well. I realized the PI folks were all using RC Astro BlurXterminator and looked into it a bit. I tried astrosharp and some others early on, but didn't like the strange artifacts. After reading some papers on U-Net and finding an example model, i fooled around with python's pytorch and got something working okay. To train it i used nasa images then added gaussian blur and added random mask noise to bloat stars. I would present the "original clean" version and the "artifical noise" version to the model for training on my old GTX 1070 GPU. Took a while, but it did start improving some images. It's not perfect by any means. To improve it I'd need to get a bigger data set and improve the star treatment by using actual plate solving on every image as currently it confuses Hii regions in small galaxies for stars sometimes etc. There are a couple few ai models to try, the "planetary" one was trained without any star treatment. ymmv but it is another tool for any experienced command line tool users. as ubersmooth is run *last* in the process at the very end, i combine it with GraXperts denoise and seems to work well. Maybe GraXpert will build in a BlurX style star treatment / deconvolution model eventually... cheers and clear skies! -john (ubergarm) |
3.53
...
·
·
1
like
|
|---|
@John Leimgruber - That sounds promising - thanks for your effort and investigation, John! As I am Linux exclusively since 1993 (30+ years) and in particular love hacking in a terminal (aka command line, bash preferred) you may understand me being keen to sneak peak the Linux version as soon as possible ...  Good progress and stay happy! CS Frank |
|
3.81
...
·
|
|---|
Hermann Klingele: This is amazing, I am amazed by the detail you've captured on the moons of Jupiter with just 8" of aperture. I will have to try your technique the next time I image the moon which has always been an old favorite. As I understand it, you capture a highly oversampled file processed using normal lucky imaging techniques, output to FITS and then deconvolution. |
|
0.00
...
·
·
1
like
|
|---|
...As I am Linux exclusively since 1993 (30+ years) and in particular love hacking in a terminal (aka command line, bash preferred) you may understand me being keen to sneak peak the Linux version as soon as possible ... lol I hear u... I did all the development on arch Linux, vim, make, and bash scripts... I've *only* been running Linux since around '95 haha... I only bothered cross-compiling a windows release so more of our club members would be able to use it. Maybe I'll clean up the sloppy code enough topush it to github for others to run and hack on. I'll keep you posted, thanks! |
|
0.00
...
·
·
1
like
|
|---|
Tony Gondola:Hermann Klingele: Yes, Siril really does a very good job here, and the key is to numerically work with the highest possible dynamic range (as if it were classical HDR photography), for both the stack and the PSF, and to ideally use a PSF from a star from the same imaging sessions, with similar exposure and stacking parameters. Here is a detail/example of a 1965 frames stack around Ganymede, PSF core with gamma 5.0 for better visibilty of structure, and the result of Siril Split Bregman deconvolution with Alpha value 24000 - without any further processing, just normalizing to ~70%, keeping some faint residual artifacts (which I normally remove for a final image):    Of course this was not an immediate result, but I really spent many hours figuring out what is important. But I found a limitation: for planets near the horizon, the OSC / RGB mode causes smearing which cannot be recovered by a PSF, at least not by one taken with higher altitude. E.g., I could not really increase quality for a Saturn imaging session compared to Registax processing. BTW, also the Edge-Rind effect on hard planet edges like on Mars gets away using such PSF! No need to complicated post-processing! |
