Processing advice please

[Ags]

This is 5 hours of data gathered on M27. My goal with the image is to capture something of the fainter outer shell. I have never gathered so much data on a target before but I am finding the processing difficult. This is the best I have managed so far.

Controlling the background stars is tricky, and getting some brightness into the outer shell tends to overexpose the inner shell.

I am using GraxPert, Siril and Gimp, I am open to using any free stuff.

I have also attached the TIFF file from GraxPert - this is the unprocessed data aside from gradient removal and denoising in GraxPert. If anyone wants to play with my data and see what they can get, that would be amazing.

Autosave_GraXpert.tiff

[TiffsAndAstro]

i had a quick go. couldn't get anywhere near as much outer nebula, sorry. there is a hint at the top but you already have that. 

[Vroobel]

I never pictured the M27, so I joined the challenge with pleasure.   Here is my vision of the object based on the delivered data. I decided to leave the background as it is, with applying a stretch, of course.  I see the Graxpert works quite well.  

PixInsight only, DynamicCrop, delicate BlurXterminator, StrarXterminator, GeneralizedHyperbolicStretch, SelectiveColourCorrection and again BlurXterminator applied to the stars as they were slightly bloated. SCNR applied to the stars (too green). 

Very good data. I'd like to work on the unmodified master file, please share it if possible. Also, could you remind us what was used? I assume the RC6, but what camera and reducer, if any?

 

 

Edited June 20 by Vroobel

[Ags]

11 minutes ago, Vroobel said:

Very good data. I'd like to work on the unmodified master file, please share it if possible. Also, could you remind us what was used? I assume the RC6, but what camera and reducer, if any?

I'll share the FITS from DSS at lunch today, thanks! 

The equipment was: RC6 reduced to 926 mm by a f6.3 SCT reducer, ZWO DUO band filter, ASI 485 MC camera, guided with a ASI120MM/FMA135 combo. Guiding on my HEM15 was around 0.7". After discarding bad frames, this is a stack of 300 1 minute subs. I am binning 2x and using gain of 200 (lower gain results in color blocks in the background).

[carastro]

I did find that quite tricky to process.  I only use photoshop and had to do a combination of Equalize function and selective layering to get this:

 

Edited June 20 by carastro

[ollypenrice]

The difference in signal between outer extensions and core is very slight, so pushing those extensions is always going to be difficult. Using a layers program (Photoshop in my case) I blended two stretches, one for extensions and one for core. To stretch the extensions I made a copy layer, added a layer mask, pasted the image onto that, Equalized it in Ps, blurred it, and stretched through that. I think Carole has done a tutorial on how to do this.

I thought the main nebula was beautiful in your data and the weakest part was the fainter stars which seemed to lack tightness and contrast. For this reason I played them down more than I usually would. Was focus nailed? My guess would be that it wasn't, quite.

In my view processing any image like this without removing and then replacing stars is on a hiding to nothing. Starnett++ is free, I think, though I used Star Xterminator.

Although I tried both DBE and SCNR Green on the data, I fould they did too much damage to the precious extensions so I didn't apply either. Very unusual for me.

Olly

[Ags]

Thanks all, before @ollypenrice's post I had already had the idea to process the image separately for each shell and then combine as layers in Gimp somehow. That seems to be what you are all doing

I am using Starnet to extract stars but obviously have a lot to learn!

Also, I think i need to shoot for double the integration time to do the outer layers justice (and I will double check that focus!).

A bigger camera would help too - framing is not ideal. It's hard to stay away from the edges with a 485MC. I am saving for a 533. Would a cooled camera really help with an image like this, or is noise now ridiculously low from modern sensors?

[Vroobel]

Imaging stars separately with an LPS filter or no filter at all is highly recommended. It's not really difficult to align the two master files according to the stars, process them separately and combine them at the end. When I work with the OSC cam I use IDAS LPS-P2 filter and do at least 30x30s frames for the stars. 

[carastro]

1 hour ago, ollypenrice said:

Carole has done a tutorial on how to do this

I had forgotten that.  All credit goes to Olly for this though as he taught me this method and we agreed I could post it on You Tube.

It should be here but for somesoI can't personally connect to You Tube at the moment, keeps tells me I am not connected to the Internet, but I clearly am or I couldnt post this.  

https://www.youtube.com/watch?v=xpPvENVdwjs 

 

Edited June 20 by carastro

[carastro]

Looks like it is because I had VPN switched on, works if I switch it off. 

[Ags]

Here is my new attempt. Star and neb stretches were completely separated. And more denoising in GraxPert...

 

Edited June 20 by Ags

[wimvb]

My version.

As @ollypenrice already noted, the outer shell is very close to the background. On a subject as this, I wouldn't use a gradient correction tool that I can't control. It is important that you can decide where background samples are placed, or the tool may destroy the outer shell. Also, some gradient correction tools (including DBE in PixInsight) will do a background neutralisation unless you disable that option. I normally do that in my images. In this image, the background was unusually dark.

I used pixinsight, but with tools which I believe are available in Siril (except blurxterminator)

spectroscopic colour calibration (any other colour calibration should give similar results)

deconvolution (BlurXtermiinator)

levels stretch and curves stretch

star removal

saturation enhancement and more curves stretch

star insertion

[Ags]

@Vroobel here is the data after DSS and before GraxPert.

Autosave.tif

[Ags]

@wimvb that is the look I am trying to get to. At least I know it's there in the data now 😀

[wimvb]

25 minutes ago, Ags said:

@wimvb that is the look I am trying to get to. At least I know it's there in the data now 😀

It's there, but you need to do a very careful stretch. In a curves stretch (not the initial stretch), put a marker just right of the histogram peak at x = y (so this point won't get stretched), then put a marker to the right of this with y > x. Put markers halfway at x = y, and a marker at x = 0.9 of maximum value and y = 0.8 of maximum value, to tone the core down a bit. Adjust the markers until you get a pleasant result.

Something like this (this is pixinsight, but any image processing software should have this function)

[wimvb]

4 hours ago, Ags said:

Would a cooled camera really help with an image like this, or is noise now ridiculously low from modern sensors?

The answer to this question used to be: cooled is better. But modern cmos is rewriting old rules. If you image all year, cooling during the summer months may be better. But during winter months, cooling may be optional. The advantage of cooling will always be that you can control the temperature. This allows you to create consistent calibration images.

[Ags]

@wimvb I think my second attempt (see above, just before your image post) is getting a bit closer, but precise stretching in gimp is not so easy, at least for me.

[Elp]

The data's definitely there:

It's not easy capturing the shells, especially after the first shell, I did it last year but did over 5 hours of O3 alone which I want to add to to get the next lot...

[Elp]

Regarding noise you generally need more time, I demonstrated the point here and this is with a DSLR:

 

The main issue I saw with the 485MC is it's lack of colour saturation especially with starfields, the 533/294/183 on the other hand you can image stars for around 30-60mins and get fine deep colour saturation.

[wimvb]

1 hour ago, Elp said:

Regarding noise you generally need more time, I demonstrated the point here and this is with a DSLR:

 

The main issue I saw with the 485MC is it's lack of colour saturation especially with starfields, the 533/294/183 on the other hand you can image stars for around 30-60mins and get fine deep colour saturation.

The 485 has small pixels and a very low full well number. This means that pixels are easily saturated, which destroys colour information. The other cameras you mention have a larger full well, and can take more photons before the pixels are saturated. To maintain colour in the stars, one needs shorter exposures and more of them.

[Elp]

28 minutes ago, wimvb said:

The 485 has small pixels and a very low full well number. This means that pixels are easily saturated, which destroys colour information. The other cameras you mention have a larger full well, and can take more photons before the pixels are saturated. To maintain colour in the stars, one needs shorter exposures and more of them.

Makes sense, the fastest I imaged with the 485mc was 10s exposures (F2), target retained colour but stars was mostly blue/white. When I did M31 at 30s exposure (F2.8) stars were mostly yellow but that could have been my colour balancing. I find those ZWO graphs misleading as they use exponential axis, but the 485 was never designed for DSO from the ground up even though it has the capability that has translated to the 585 pro.

[ollypenrice]

3 hours ago, wimvb said:

It's there, but you need to do a very careful stretch. In a curves stretch (not the initial stretch), put a marker just right of the histogram peak at x = y (so this point won't get stretched), then put a marker to the right of this with y > x. Put markers halfway at x = y, and a marker at x = 0.9 of maximum value and y = 0.8 of maximum value, to tone the core down a bit. Adjust the markers until you get a pleasant result.

Something like this (this is pixinsight, but any image processing software should have this function)

Yes. In processing, the stretch is everything. You're showing a single stretch (with the stars awaiting a different one) but sometimes different stretches in different layers may be necessary. But... the stretch remains everything. That's why I don't use ready made stretches. They are guesses, sometimes clever ones, but still guesses.

Olly

 

[wimvb]

1 hour ago, ollypenrice said:

That's why I don't use ready made stretches

Me neither. Any youtube video or other tutorial, that promotes using the STF function in PixInsight as a first stretch, gets a big minus on my list. I’ve also never had any luck with the new GHS stretch that people are raving about, except on the Orion nebula, when I followed Adam Block’s video. I prefer connecting dots in a curves stretch, or moving a mid tones slider.

[wimvb]

2 hours ago, ollypenrice said:

You're showing a single stretch (with the stars awaiting a different one) but sometimes different stretches in different layers may be necessary.

As you know, pixinsight (the spanish inquisition, although they’re not so frantic anymore, it seems) doesn’t do layers. I stretch most of my images without masks, and my wide field galaxy images with stars in place. Understanding stretches is a must in this hobby.

Edited June 20 by wimvb

[Ags]

OK, here is take three. I did three layers - one for the out shells, one for the core, and one for the stars. In terms of progress, there are less stars and those that are there are less ugly. The outer shell is more defined and smoother, and there is more definition and color in the core.