Reprocessing the Lagoon and Trifid nebulae in Siril.

[Felias]

I have not been able to go out with my telescope for weeks, so I am practising with old photos to improve my processing skill. This June picture, almost ruined by too much dew on the lens...

...has turned into this:

The first picture was stacked in DSS and processed in PS. The new, less cropped version was binned and stacked in Sequator, then processed with Siril. I removed the stars with Starnet++, then merged and denoised in PS. Both 170 x30s subs, ISO 400. WO Z61II, unmodded Canon 77D.

I think it looks better, especially thanks to the photometric calibration in Siril. Maybe I should have processed less the centre of the Lagoon, left it brighter? I'd appreciate any tips on how to improve the result (apart from getting more data, I mean).

[Dazzyt66]

Wowsers! I’d be happy with that as is! I use Siril for processing too and have found an initial Sinh stretch (after background extraction, colour cal etc) helps with detail before going on to further stretching. I’ve just started using Startools and that has given me some promising results too on nebulae  (using Siril for stacking) 🙂

[Felias]

Thanks. Yes, I do a sinh stretch before I start with the histogram in Siril. It's not worth stacking in it for me, though, because it takes more than a day using all the RAM in my laptop, rendering it unusable. I could leave it overnight, but in any case I wanted to bin the noisy data, so Sequator did the job. Can you bin in Siril?

I tried startools months ago and didn't get good results. Maybe I should try again, now that I know more about processing.

[Dazzyt66]

I’m not sure if you can bin in Siril as I usually do that in Startools. I found with ST the first time I used it I was starting off wrong - a few YouTube views later I’m getting better results 😊

Edited October 3, 2021 by Dazzyt66

[alacant]

3 hours ago, Dazzyt66 said:

not sure if you can bin in Siril

There's quite a good 're-sample' under 'Geometry' but it seems to lose the faint bits which StarTools' 'Bin' retains.  

What we haven't tried is a Siril re-sample followed by subsequent processing in StarTools. Maybe then, the outcome would be similar... ?

Edited October 3, 2021 by alacant

[Felias]

On 03/10/2021 at 12:00, Dazzyt66 said:

I’m not sure if you can bin in Siril as I usually do that in Startools. I found with ST the first time I used it I was starting off wrong - a few YouTube views later I’m getting better results 😊

 

On 03/10/2021 at 15:20, alacant said:

There's quite a good 're-sample' under 'Geometry' but it seems to lose the faint bits which StarTools' 'Bin' retains.  

What we haven't tried is a Siril re-sample followed by subsequent processing in StarTools. Maybe then, the outcome would be similar... ?

 

Thank you, I guess the case for Startools gets stronger... 🤔

[Dazzyt66]

8 hours ago, Felias said:

 

 

Thank you, I guess the case for Startools gets stronger... 🤔

You can try it for free and see how you fare. In my case it took several work throughs of one of my images until I'd achieved something better - mainly to do with the initial import and manipulation in ST.

[alacant]

59 minutes ago, Dazzyt66 said:

mainly to do with the initial import and manipulation in ST.

Hi

To make sure you have your data in the best shape possible (we need every gram we can muster) be sure to read the dataset preparation detail.

Cheers and good luck

[Dazzyt66]

8 hours ago, alacant said:

Hi

To make sure you have your data in the best shape possible (we need every gram we can muster) be sure to read the dataset preparation detail.

Cheers and good luck

This is what I follow albeit for Siril rather than DSS for prep. The main problem (apart from user error 😂) was which ‘linear’ selection to make initially in ST as that makes a massive difference to image quality. The colour section will also be down to personal preference as I personally don’t like the default apply. It’s pretty straight forward once you work out your workflow and worth the money if you don’t want to spend PI or PS wise. 😊
 

ps I’ve tried doing some initial processing in Siril and then moving across to ST with not much success so I find it best just to stack and then go to ST - again that could be user error!

Edited October 6, 2021 by Dazzyt66

[alacant]

2 hours ago, Dazzyt66 said:

best just to stack and then go to ST

Exactly as recommended. Do nothing in Siril apart from calibrate and stack.

 

[Felias]

On 06/10/2021 at 09:03, alacant said:

Hi

To make sure you have your data in the best shape possible (we need every gram we can muster) be sure to read the dataset preparation detail.

Cheers and good luck

I did exactly that the first time around, but only ended up with a green blotch in the end. I still have Startools in my laptop, so I will try again. The stacking will be in DSS or Sequator, though, I can't afford the 15+ hours that Siril takes, not to mention that the computer gets so hot it may melt down! When you say calibrate in Siril, what do you mean exactly? Is it the colour?

[Dazzyt66]

9 hours ago, Felias said:

I did exactly that the first time around, but only ended up with a green blotch in the end. I still have Startools in my laptop, so I will try again. The stacking will be in DSS or Sequator, though, I can't afford the 15+ hours that Siril takes, not to mention that the computer gets so hot it may melt down! When you say calibrate in Siril, what do you mean exactly? Is it the colour?

I’ve started to use DSS too to check the difference - I think DSS does get the final stack data slightly better for ST. I followed the exact setup as per the link. The thing that makes the difference in ST at the start is if I select ‘linear’ I get a very bright green image until first wipe - which isn’t useful for seeing what needs fixing, so as you are using a DSLR make sure you use the ‘linear for DSLR or OSC….’ selection. One thing I have found (and I think it’s just my ST skill atm) is that ST is better for nebulae etc whereas I can get a better final image just using Siril/Gimp for Galaxies. I haven’t had any success (so far) if I additionally process the post-stack more in Siril in any way before opening in ST.

Edited October 9, 2021 by Dazzyt66

[alacant]

15 hours ago, Felias said:

calibrate in Siril, what do you mean

The same stuff you do with the bias and flat frames in dss before you stack. I'd recommend both calibration and stacking in Siril.

Edited October 9, 2021 by alacant

[alacant]

10 hours ago, Dazzyt66 said:

I haven’t had any success (so far) if I additionally process the post-stack more in Siril in any way before opening in ST.

ST works best with linear data. Anything else you do before opening the new stack in ST will deteriorate the results obtained therein.

Cheers.

[Felias]

11 hours ago, alacant said:

The same stuff you do with the bias and flat frames in dss before you stack. I'd recommend both calibration and stacking in Siril.

I see, thanks. I'd love to do the stacking in Siril, but as I said it's too time-consuming and I'm afraid it may damage the laptop, it just gets too hot for more than half a day. I'll try DSS and Startools again, and see if this time around I can improve my previous attempt!

[alacant]

1 hour ago, Felias said:

too hot for more than half a day.

On a standard install of Ubuntu 20.04, even on a modest i5, it's fast. Are you sure you have the latest version with all the cpu threads in use?

I'm sure that this would be if interest to the developers who are sure to suggest a fix.

Cheers

 

Edited October 9, 2021 by alacant

[Felias]

3 hours ago, alacant said:

On a standard install of Ubuntu 20.04, even on a modest i5, it's fast. Are you sure you have the latest version with all the cpu threads in use?

I'm sure that this would be if interest to the developers who are sure to suggest a fix.

Cheers

 

Yes, I've got the last version, and my laptot is a Lenovo Legion, i7 and 32 GB of RAM. Windows 10, though, so that may be the issue. Anyway, I've just tested Startools with my data for M31, which I had readily available, and I'm finding difficulties again. This is my previous Siril/PS processing:

 

 

And this is what I did in Startools (a screen capture, so low res):

I was quite frustrated when I tried to use HDR to bring detail, I couldn't use the star mask and they became bloated with a sharp centre, absolutely awful. I had to settle for the adjustment to enhance the core of the galaxy, but it came out too soft and dark, I prefer the core to be brighter as in the first picture. Then the colours were difficult to adjust, and there was a lot of guessing involved, whereas the photometric calibration is Siril is quite straightforward. And in the end, Startools didn't recover much more nebulosity than Siril did, it seems, so it's not a trade off I'm sure that works for me. Obviously, I can't tell just by this processing alone, I'll keep practising and see if I can improve...

[Dazzyt66]

Either way these are great images (although I do prefer the first). On your ST image how much noise removal did you use? - I've found if I go much higher than it suggests that I get too blurry final image -which I think may be the case here - detail wise it looks very similar to the other version.

I have also found that I get a better (to my eyes) version of M31 in Siril using colour cal etc. and final tweaks in GIMP than I do in ST. Although to be fair, the more I practice with ST the better they are getting. If I do 'nebulous' images such as NGC7000 or IC1340 they come out really well with just presets in ST. Here's what I mean about M31, first i Siril and Gimp and second from ST (my latest try which is probably version 10!)

 

 

 

I run Siril under windows 10 on my Toshiba laptop which is i5 7200 2.5ghz with 8gb ram all OK - in fact its similar speed to DSS process wise...

Edited October 10, 2021 by Dazzyt66

[ollypenrice]

A simple trick which is particularly effective with OSC data on Ha-rich targets is to use Photoshop's Selective Colour (Image-Adjustments-Selective Colour)) to lower the cyans in red. This brings up the Ha signal remarkably well. Here I just used it on a screen grab but it would be far better, of course in an uncompressed format.

If you want to get clever you can use the modified version as a luminance layer over the original, allowing you to see more faint Ha without changing the colour, or you can apply it fully as luminance and paritally as colour, or whatever.

After de-starring in Starnet, how did you re-apply the stars? I tend to do it in Ps by opening the starless version, pasting the top layer in linear form on top, and changing the blend mode to Lighten. At first the top layer will then be entirely invisible, but if you give it a simple log stretch in Levels the stars will begin to appear. You can stop stretching at any time, meaning you can keep the stars at any size you like.

Olly 

 

Edited October 10, 2021 by ollypenrice
clarification

[alacant]

9 hours ago, Felias said:

Startools didn't recover much more nebulosity than Siril

Surprising. Post a link to the data if you like.

I'd put good money on the second -and any subsequent AutoDevs- being taken with non-optimal ROIs though.

Excellent images anyway. Well done.

Cheers

Edited October 10, 2021 by alacant

[Felias]

result.fit

9 hours ago, Dazzyt66 said:

Either way these are great images (although I do prefer the first). On your ST image how much noise removal did you use? - I've found if I go much higher than it suggests that I get too blurry final image -which I think may be the case here - detail wise it looks very similar to the other version.

 

Thank you! Yes, there is perhaps too much noise removal in the Startools version, but since I was doing just a screen capture I didn't bother refining that part. I think it was the "life" tool that applied the denosing automatically. Not sure now.

7 hours ago, ollypenrice said:

A simple trick which is particularly effective with OSC data on Ha-rich targets is to use Photoshop's Selective Colour (Image-Adjustments-Selective Colour)) to lower the cyans in red. This brings up the Ha signal remarkably well. Here I just used it on a screen grab but it would be far better, of course in an uncompressed format.

 

If you want to get clever you can use the modified version as a luminance layer over the original, allowing you to see more faint Ha without changing the colour, or you can apply it fully as luminance and paritally as colour, or whatever.

After de-starring in Starnet, how did you re-apply the stars? I tend to do it in Ps by opening the starless version, pasting the top layer in linear form on top, and changing the blend mode to Lighten. At first the top layer will then be entirely invisible, but if you give it a simple log stretch in Levels the stars will begin to appear. You can stop stretching at any time, meaning you can keep the stars at any size you like.

Olly 

 

Yes, I know about selective colour, I did exactly that with the reds in PS. So that means yours is a second iteration of the same process! 😉 I'll try the luminance layer, I didn't think about that, thanks. As for Starnet, I don't like the way the starless version looks, so I just use it to build a good star mask in PS. Then I use the 'minimum' filter, trying not to push it too much (I do like a good star field). I've also tried to use the starless layer to do a 'content aware' fill in PS, it produces good results sometimes. I'll try your suggestion.

5 hours ago, alacant said:

Surprising. Post a link to the data if you like.

I'd put good money on the second -and any subsequent AutoDevs- being taken with non-optimal ROIs though.

Excellent images anyway. Well done.

Cheers

I attach the stacked file, thank you. I played with different RoI, but it wasn't great (is there a way to select one diagonally?). There is a big black smudge in the upper right corner (dirt on the lens) that didn't help, but found that a carefully chosen mask kept along the process did wonders. I'd love to see what someone experienced in Startools can do with the data! 🙏

result.fit result.fit

[Dazzyt66]

Here's my go. Nothing fancy, just what I've learned over last few days - you've got some great data there - I'm hopeful I can get something like that at some point! Did a .tif version so you can save and alter colors as you wish - I think there's a bit too much green in there atm... I would normally do some further colour tweaks in Gimp but this is pure ST output...

M31ST.tif

Edited October 10, 2021 by Dazzyt66

[alacant]

2 hours ago, Felias said:

love to see what someone experienced in Startools can do with the data!

So would I:)

This is my go, developed -sorry, couldn't resist- to within a fraction of a milimeter of its life.

There is some lovely detail. But yes, I see what you mean about the dark anomaly.

Cheers

[Felias]

1 hour ago, Dazzyt66 said:

Here's my go. Nothing fancy, just what I've learned over last few days - you've got some great data there - I'm hopeful I can get something like that at some point! Did a .tif version so you can save and alter colors as you wish - I think there's a bit too much green in there atm... I would normally do some further colour tweaks in Gimp but this is pure ST output...

M31ST.tif 16.86 MB · 0 downloads

 

Thanks! It looks very green on my screen, though I know my laptop has issues with the colour calibration. And the blacks look clipped, I'd say. I've done a second attempt this afternoon, it is better, but I'm very frustrated that I can't use a mask in modules such as contrast and HDR. I could always enhance in PS, but then I'd lose the denosing made in Startools. I didn't bother cleaning the dark smudge in the corner, for a screen capture, so here it is:

 

[Felias]

7 minutes ago, alacant said:

So would I:)

This is my go, developed -sorry, couldn't resist- to within a fraction of a milimeter of its life.

There is some lovely detail. But yes, I see what you mean about the dark anomaly.

Cheers

Thank you! Your image has the same problem I encountered: if I want the core to remain bright, the stars get a sharp centre with some halos that ruin the image for me. That's why I had to use the 'reveal core' option, which leaves the stars untouched but looks less realistic (I think the core should be bright, as in your picture, considering the large star density in the region). Is there a way around this? Also, the 'contrast' module can't be used correctly as long as the smudge cannot be masked, it adds an annoying gradient.