First go at M81 and M82

[SkyExplorer]

Thanks, so first time around I am looking for sample points across the whole image (apart from foreground items)?

[SkyExplorer]

This is the image without crop or sncr, just an auto STF

[glowingturnip]

do a crop just to get rid of the very edge stacking artifacts, otherwise they'll throw the auto-stretch off.

I think you've got your channels linked on the auto-stretch which is why it's coming out green - open the ScreenTransferTool (STF), make sure the link RGB channels button (top left button) is unchecked, then hit the button that looks like a nuclear warning sign.  Should give you a better neutral-colour STF.

Thanks, so first time around I am looking for sample points across the whole image (apart from foreground items)?

yep, and make sure the crosses are not red - increase the tolerance parameter if they are

[SkyExplorer]

Thanks Stuart,

After a crop, then running the CanonBandingReduction script, followed by a DBE, then another DBE with Normalise selected I have this.

This is the background from the second DBE, looks very pretty :-)

[SkyExplorer]

Just removed some of the green using SCNR, what should be next step ?  

It seems to be looking a bit better now, I think I was over stretching earlier today and getting a bit disheartened at the amount of noise present.

[SkyExplorer]

Carried out a Colour Calibration as per Harry's galaxy method.

[glowingturnip]

starting to look good now, isn't it 

quite a bit of noise there (don't worry, I'm a fellow DSLR user and get more than my fair share of noise).  Check out the noise reduction in MultiscaleLinearTransform and in TGVDenoise, i think there should be tutorials on the web somewhere for them (if not, shout).

[gnomus]

Carried out a Colour Calibration as per Harry's galaxy method.

There ya go.......

[SkyExplorer]

Thanks guys, will have a look for those tutorials tomorrow. Thanks again.

[SkyExplorer]

Found a tutorial on TGVDeNoise and ACDNR, so tried those, then followed Harry's HDR video followed by his "Local Histogram Equalisation" video to bring back some contrast.

Is the redish background normal, or is it the effects of LP/sensor noise or not enough DBE's?

[glowingturnip]

that's way too much noise reduction I'm afraid, giving you that blotchy over-smoothed look.  You should be aiming at just enough/not quite enough noise reduction to keep it still looking natural - a little goes a long way.

I like MLT (Multiscale Linear Transform) noise reduction and usually apply that first, followed by some light TGVdenoise.  I don't bother with ACDNR, I think TGV has pretty much replaced it.

Here's how I do my MLT:

- set a preview of a small area (about an inch square with some background and some detail in it, so you can see the noise close up)

- launch MLT, set it to 6 layers, and turn off layers 1-5

- launch the real-time preview window on your previewed area - it will just look like a big blur because you've got all those layers turned off

- turn layer 1 back on - now you'll see the noise attributed to that layer pop into view - it's kind of like looking at a ground glass screen

- tick the noise reduction for that layer, set the amount to 0.75, the iterations to 3, and then reduce the threshold to the point where you can see the noise just wink in and out of view, set it just at that point, no more

- turn off layer 1, and then do the same for layers 2, 3 and 4 in turn.  3 and 4 you should be very light with the noise reduction, usually no more than about 0.5

- close the real-time-preview, turn all the layers on, and apply !

TGV you need to be very light-handed with - not enough is much better than too much.

Hope this is useful ?

[SkyExplorer]

Thanks for the tips Stuart,

When I setup MLT and make layer 1 visible, I see some grey halos around the stars I have in my preview (a bit of M81 and background)  When I reduce the threshold from the default of 3, the grey halos get worse.  Should I increase until they disappear  ?

[glowingturnip]

no, concentrate on the speckles of noise, reduce until they're just disappearing.

the halos round the stars with the other layers turned off are just an MLT thing, don't try to minimise them

[frugal]

Hope this is useful ?

That is one of the most useful PixInsight tips I have heard in a while. I had not realised that if you disable a wavelet layer it removes it from the result entirely. I had always assumed it just meant "don't do any processing on this layer". That is going to make all kinds of multi-layer processing easier.

[SkyExplorer]

This is the linear image auto stretched after doing a few DBE's on it (showing preview for MLT) followed by the MLT image.  M81 seems to have lost some detail after MLT. I did find a tutorial on MLT that used a mask for the layers ( http://pixinsight.com/examples/M81M82/ ) so that MLT ignored the galaxies, but not sure if you apply a mask for every layer.

A quote from the article "Once we have found good linear mask parameters—which is much easier than what may seem at first sight, since the tool guides the process"...  Err, not sure if the tool guides the process.

After MLT

[glowingturnip]

not sure if you have lost detail or not, I think I can still pick out most salient features, but if you think you have, you can always go back to the MLT tool and reduce the threshholds a bit.  You can use a lightness mask for the background too, I tend not to because I think it looks a little odd when you have more grain showing on the galaxies than on the background, but if you do it carefully it can work.

TGV wil be able to target those dark spots that MLT can't quite get to, but use it carefully.

your auto-stretch is starting to give you a wildly horrid background now, isn't it.  To see it properly, you could a) load up a previous version with a nicer stretch on it, maybe from post 31, auto stretch that, then right click on it, screentransferfunction/applystretchtoallopenimages or open the screen transfer function tool and adjust the sliders manually.  You're about getting to the stage where you can do a non-linear stretch with the Histogram Tool though.

[SkyExplorer]

I used the Histogram Transfer Tool and created this.  M81 looks a bit odd though, a bit fuzzy.

[glowingturnip]

yeah it does a bit.  maybe dial back on the noise reduction a bit then.  Also i think you're black clipping it a bit, black slider on the histogram too high, which might be losing some of the fainter detail on M82.

Is this with HDR Multiscale and LHR Equalisation applied, or not ?

M82 is starting to look cracking though.

[SkyExplorer]

Yes both applied.

I'm really happy with M82.  It's hard as M81 is a lot fainter than M82.

Do you mean less TGV or less MLT?

When working with the Linear image is it best to use Auto stretch and leave it, or adjust it so the background is not as harsh, i.e so it looks like post 31?

[glowingturnip]

at the risk of creating too much work, what would it look like without HDR and LHE, but with the DBE and noise reduction you've done so far ?  Hopefully you've got the steps/process icons saved ? 

I usually muck about with those two after all the other steps you've done and after the initial histogram stretch

[SkyExplorer]

This is the image before and NR.  I have done DBE, Crop, SCNR and DBE again (as per the image title).

After NR it looks like post 40.

[SkyExplorer]

Stuart,

I'm guessing when using MLT I need to change the Algorithm from the default of "Starlet Transform" to "MLT" ?

When using MLT I don;t really see any noise in layer's 1 and 2.  Lots in layers 3,4 and 5.

[SkyExplorer]

I was not sure about the amount of layers, as you mentioned 6, then turn off 1-5, but you did not mention turning on layer 5 and doing NR on that.  So do I need layer 5 to stay off and 6  to stay on without NR?

I thought I show you a screen shot of the MLT windows as it may help :-)

BTW, thank you for your help with this, I am very grateful 

[glowingturnip]

BTW, thank you for your help with this, I am very grateful 

no problem, just hope I'm not boring everyone silly !

Keep the algorithm as Starlet Transform - it's actually the Pixinsight equivalent of wavelets, and used to be called Atrous Wavelet Transform (if you were to increase the bias parameter, it sharpens in just the same way that Registax sharpens, but here we're applying denoising on a layer by layer basis instead).  Multiscale Transform is a new one from Pixinsight, and I'm sure has a lot of science behind it, but I've not had decent results with it yet tbh.

I don't use denoiseing on layer 5, and only sparingly on level 4, as I find they start to leave artifacts around genuine structures.  I turn it off for the real-time preview just to give a view without detail in it so that the only thing apparent is the noise in each layer.

[SkyExplorer]

Thanks Stuart,

The image I posted in 48, is that preview good or bad in your eyes?