StarTools vs Siril - A challenge

[ajh499]

I recently posted an image processed in StarTools on the Reddit astrophotography sub - https://www.reddit.com/r/astrophotograp ... te_nebula/

A user over there suggested that I might get better results these days with Siril. After a bit of back and forth, I sent him my data and he had a go at processing it and got the following image using only the LRGB (no HA)

fin.jpg (350.47 KiB) Viewed 2211 times

It has all sorts of issues and artefacts (it was only a quick reprocess attempt), but there is detail there that I just cannot get StarTools to pull out. Plus I've always had issues with star sizes in StarTools and these look much better too

Can anyone here do a better job than I can with my data? https://www.dropbox.com/scl/fi/k25pp5s0 ... 87f8x&dl=0

A full size version of the Siril image is here https://www.dropbox.com/scl/fi/kuprvrzj ... y4fs5&dl=0

I've used StarTools for years, but for a number of reasons (not simply this image) I'm really considering switching to Siril. Can anyone persuade me to stay

Alex

[dx_ron]

I have no reason to try to persuade you to choose one program over another, though the Siril version Rob produced for you is just objectively bad. He only used the LRGB, yet produced colors that have no business being there. So if you're after that sort of artistic impressionism, then great. I assume most of the gaudy colors came from Affinity? I don't think Siril would make anything like that. I also rather doubt that Siril would give you so much "detail" - Rob's version has been sharpened way past any rationale point.

Here's a link to a nice reference for what and RGB-only (or LRGB) Rosette ought to look like: https://www.astrobin.com/oxwfy3/0/ (just a nice one that I found on astrobin)

Here is a 100% view of Rob's version:

100pct.jpg (333.5 KiB) Viewed 1895 times

I did notice with your data in ST that maybe focus wasn't great. Pre-SVD dim stars were not particularly star-like:

before_svd.jpg (54.87 KiB) Viewed 1895 times

they became much more point-like after SVD:

after_svd.jpg (61.91 KiB) Viewed 1895 times

continued - because of the forum software's 3 attachment-per-post limit

[dx_ron]

I make no claim that my ST version (using the Ha data) is the best possible (or even all that good) - there are plenty of people way better at this than I am, and I didn't spend too much time on it. I did no sharpening (because I'd already fiddled with masks long enough), but I did some un-glow in Shrink wit some (not all) of the stars. Not great, more work would be required to make it look less artificial.

Overall, your data are noisy - and it is true that Siril can often produce images from noisy data that are more pleasing to the eye - for reasons I do not understand. It is also true, obviously, that removing stars and re-introducing them can make for smaller stars. So if that's the style you prefer - go for it!

NewComposite.jpg (488.92 KiB) Viewed 1889 times

[ajh499]

Ron, thanks for taking a look

He only used the LRGB, yet produced colors that have no business being there.

Yes, I agree, there's a particularly "abstract" selection of colours there. I wasn't suggesting atempting to emulate that, just I wish that I could get more "depth" in images using StarTools. All my attempts, like yours, look artificial or pasted on, and with stars that look hazy.


The data is fairly noisy, about 50% of the exposure time is when the Rosette was at less than about 30 degrees elevation so that's a but unaviodable.

I'm not sure why the stars are quite so soft, I was autofocussing like crazy while capturing so it should be as sharp as I can get the focus, maybe I haven't discarded enough dodgy subs, it was windy - 45mph gusts do make guiding tricky! But this image doesn't look all that different from any others I take, either with this ED80 or with the FMA135. Maybe I'm doing something wrong optically.

I like the philosophy of StarTools, I want naturally coloured images (as much as possible), I don't want gaudy fake images with artifacts everywhere and weird colours, but I'm rarely happy with what I end up with.

I don't know what I'm doing with Siril, but I managed to get the following image just from the LRGB (no Ha). Noisy data aside, I don't think it's too far from that Astrobin image you linked to, but I doubt I could do anything similar with StarTools and that seems a shame.

rosette.jpg (277.96 KiB) Viewed 1636 times

[Stefan B]

Hi Alex,

I had a quick look at your data but have to admit that I ended nowhere near your last rendition done in Siril which looks really nice in my book.

But I noticed that your luminance layer contained some stacking artifacts due to frame rotation I guess:

stretch.jpg (288.05 KiB) Viewed 1589 times

In my experience ST hates this stuff. Wipe and stretch gets thrown off. Maybe that's part of the problem. On the other hand cropping away the corresponding areas didn't help much with getting a better result.

Regards
Stefan

[ajh499]

Stefan

Thanks for taking a look

Yeah, I forgot to mention that the corners need cropping out, I've done that in all my versions of this image both with StarTools and with Siril. It's due to the fact that I forgot to check rotation between two nights.

Alex

[dx_ron]

Played around some more (because I want to improve my own processing and this seemed an instructive challenge).

First thing I did differently was to bin quite a lot right from the start - all the way down to 1600px wide (because that's the final size I would use for posting to a forum anyway). That seemed to really help Wipe, both the L channel and especially the chrominance. Then I cheated a bit for Alex - I masked the brightest stars and used Layer to un-do deconvolution - hopefully leaving more Siril-esque large stars.

In the Color module I pulled bright saturation way down and dark saturation all the way up (the reverse of the default). Then I hit it with a small amount of Saturate in SS. After keeping that, I went back into SS and played around. I've never really tried the Brighten button, but I actually quite like what it did for this image, so I kept that. A bit of work in Shrink put me here:

rgb_2.jpg (470.06 KiB) Viewed 1559 times

Which I think is much more like your Siril coloration.

I went then to NBaccents. You have to realize that you only collected Ha, so it is inherently going to "flatten" the image a bit because it can only increase red (or purplish with the Balmer setting). I dialed the strength back to maybe 45% (I think it was) to keep the impact down. Here's the "plus a bit of Ha" version:

Ha-rgb_1.jpg (472.55 KiB) Viewed 1559 times

[ajh499]

Ron

Oh that looks good. I'll have to give it a try using your processing steps. Would you mind posting your log of this session?

I didn't think Wipe was having much of a problem with this image, but I binned to 50% as a matter of course anyway, so maybe I never saw an issue.

Undoing the decon on the bright stars is a clever idea, I'll have to give that a try. It's not really that I prefer larger fluffier stars, it's just that I really don't like the single sharp point with a big coloured halo around it that I always seem to get with StarTools. It makes the image look like it was shot through fog or haze.

Thanks


Alex

[admin]

Hmmm....

I'm just not seeing how, given the rather noisy and limited LRGB dataset provided, one can arrive at any of the Siril images without some heavy (and I mean heavy!) detail synthesis.

Can you confirm neither Siril images had any detail synthesis involved? To me, both images seem heavily doctored (missing stars, stars reinterpreted as irregular planetary nebulas or gas knots, many structures that don't exist in other, more high-res renditions, etc.) while showing detail/sharpness beyond what a single star's PSF in the dataset otherwise suggests should be possible for your scope and atmospheric conditions at the time.

StarTools is a (documentary) astrophotography package, and was never meant to cater to art, make-belief or fantasy renderings. It stays true to your data. If you have developed a desire to go beyond your data/reality, I would indeed suggest StarTools is no longer the right software for you. While I would find it sad to see you go, I would be even sadder about the reasons why.

I fully understand it is now easier than ever to become discouraged in the face of the many (we're closing in to the majority these days IMO) deep-faked images that are going around, but you have to ask yourself whether you are in it for the bragging rights and internet points, or for discovering and documenting truth and reality.

It is incredibly discouraging how far the quality of images have slid deep into make-belief territory, where seemingly anyone with an iPhone is now producing Hubble-like images, all the while patting each other on the back and pretending it is all great and (somehow) more advanced.

Compare any of these images to each other, or to images from a few years back, and it quickly becomes clear that everyone is now simply picking their own reality (which is not the same as interpretation of a single, shared reality!).

Whereas a few years ago we would compliment each other on managing to capture and bring out this or that hard-to-capture feature, today many astrophotographers don't even know or (care to?) understand these features. Recently I saw someone "accidentally" remove the remnant star that is the progenitor of the planetary nebula they were capturing. No one bat an eyelid. It made me realise how severely astrophotography is not in a good place.

[ajh499]

Ivo

Thanks for the reply.

I'm just not seeing how, given the rather noisy and limited LRGB dataset provided, one can arrive at any of the Siril images without some heavy (and I mean heavy!) detail synthesis

I can't speak for the first Siril image, I didn't create it, but there's obviously all sorts of crazy business going on there. I wouldn't look too closely at that image as, I agree, there's something very odd happening there.

My Siril image had nothing done to it that should have synthesised anything. I really only removed the gradient, stretched and deconvolved it.
The only artefacts will have come from the fact that I used Starnet to separate the stars and nebula and stretched them separately. Starnet left some artifacts behind where it didn't completely remove the stars. The most obvious example of that is on a star in the bottom left corner, but there are a few others too.

Screenshot 2025-01-01 133527.png (87.73 KiB) Viewed 1265 times

StarTools is a (documentary) astrophotography package, and was never meant to cater to art, make-belief or fantasy renderings. It stays true to your data. If you have developed a desire to go beyond your data/reality, I would indeed suggest StarTools is no longer the right software for you

As I said in an earlier post, I like StarTools philosophy - I agree with you, and while I'm not aiming to produce scientifically accurate images, I will always err on the side of staying accurate as much as possible. I don't want fantasy or AI produced images either.

My recent disappointment with my StarTools images have been due to a couple of things that I have never been able to resolve.
First, stars look "hazy" when using AutoDev - the centre coaleces nicely into a point, but leaves a coloured halo that makes it look like we're seeing through fog.
Nubulosity often seems "crunchy" or maybe "quantized", I'm not sure how to explain it, but it feels unnatural and sometimes like it's pasted on. My Reddit version and Ron's first image look like that too. His second image is much better, but still has a slightly artificial look to it, maybe in that case it's just a bit too much contrast. Maybe that's my issue too.

Following Ron's suggeston to undo the decon of the brightest stars, I had another go with StarTools and got the following. It's not as good as his for star tightness and I don't think it gives the impression of a cloud the way that my Siril version does, but it's a bit better. If only I could sort the stars out

Rosette2Nights-LRGBHa-3.jpg (525.97 KiB) Viewed 1265 times