M51 LRGB

[Datalord]

Got around to processing about 11 hours of LRGB on M51. Feedback very much appreciated.

Gear:   CFF12"RC, G3-16200, Celestron CGX, Baaders LRGB filters

Data:   L: 34*600s  R: 31*180s  G:31*180s  B:28*180s

Process: PI stack, Drizzle lum, DBE, decon lum, MLT noise and sharp, combine, background neutralization, SCNR, ColourCalibration, morph, curves. PS selective sharpening and noise reduction

[Seanelly]

Super image.

[Barry-Wilson]

Hi Thomas

If you are ok with feedback as you have requested, some questions arising in my mind:-

I think you have a nice flat background sky and some good colour in your star field, neither of which appear over processed.

I would hazard that your data contains more detail than the final image and that perhaps you sense this with your request for feedback?

I wonder if you have lost detail in your image at your deconvolution as I see you have tell tale 'little worm' shapes.  Combining your luminance with your RGB has left it washed out.  Maybe incorrect NR settings have reduced detail in the galaxy too.

Why did you decide to drizzle the luminance?

Is your list of PI steps in the order that you processed?  If so, there are certainly steps in the wrong order, eg colour calibration after linear stretching.

What chronological workflow steps did you use?

Is this image cropped?  What is your native image scale with your kit?

I'm happy to help with processing steps, just let me know.

HTH

[Datalord]

1 hour ago, Barry-Wilson said:

If you are ok with feedback as you have requested, some questions arising in my mind:-

Absolutely, I hope to learn.

1 hour ago, Barry-Wilson said:

I wonder if you have lost detail in your image at your deconvolution as I see you have tell tale 'little worm' shapes. 

I thought I had dialed it back so far that it enhanced the overall image, but maybe I still miss the perfect settings. I start with the lightvortex tutorials and have experimented from there.

1 hour ago, Barry-Wilson said:

Maybe incorrect NR settings have reduced detail in the galaxy too

This is actually a strong possibility. I used MLT with different settings for light and dark areas in lum.

This is my master lum before LRGBCombination. Looks like I lose detail after combining.

1 hour ago, Barry-Wilson said:

Why did you decide to drizzle the luminance?

I shot this with bin2 at 1.05"pp, so drizzle brings the resolution up a bit. I guess I should reconsider, as I just took a side by side of the clean stack next to the drizzled:

seems like the non-drizzled has more detail. Both are nuked with STF.

1 hour ago, Barry-Wilson said:

Is your list of PI steps in the order that you processed?  If so, there are certainly steps in the wrong order, eg colour calibration after linear stretching.

pretty much. I thought I had to calibrate before stretch.

1 hour ago, Barry-Wilson said:

What chronological workflow steps did you use?

pretty much this. I didn't do decon and MLT on the colour master, except for MLT noisereduction on the galaxy as I read somewhere the colour is less important in sharpness. 

1 hour ago, Barry-Wilson said:

Is this image cropped?  What is your native image scale with your kit?

Yeah, here's full lum in one-third. 

1 hour ago, Barry-Wilson said:

I'm happy to help with processing steps, just let me know.

Help much appreciated. No reason to do all this work half arsed.

[Barry-Wilson]

Thanks for that info.

Re: colour calibration - you do carry out this step before stretching, from your original thread it's just that you mentioned it towards the end of the list.

Modifications I would suggest from reading your screenshot of the PI processes (but I may guess the order incorrectly):

1. As you have seen, I wouldn't carry out the drizzle step.

2. Carry out SCNR after your RGB stretch in the non-linear stage.

3. Carry out your crop after RGB combination (just saves you some steps).

4. Carry out DBE on the combined RGB image rather than on separate channels (saves steps and you can properly see the colour gradients).

5. Do not use LRGB combination - it washes out the colour - CieLab combination with masks is the way to go.  Please see my tutorials on basic LRGB workflow, including DeConvolution, here.

6. Definitely check your DeConv settings or reduce the iterations as you are creating artefacts (worms) and losing detail.

6. NR and sharpening after stretching in the non-linear stage.

HTH - more techniques in my tutorials too.

 

 

[Barry-Wilson]

Sorry if my reply above seems a little brusque, I'm just dashing about this evening trying to get jobs done ? ?.

[Datalord]

41 minutes ago, Barry-Wilson said:

Carry out your crop after RGB combination (just saves you some steps).

The benefit of saving the crop and doing it for all at once is that I can reuse masks and DBE processes across. also

43 minutes ago, Barry-Wilson said:

4. Carry out DBE on the combined RGB image rather than on separate channels (saves steps and you can properly see the colour gradients)

I'll try, but my experience is that the gradient will be slightly different between the colours and only by individually running DBE can I get a clean colour master.

44 minutes ago, Barry-Wilson said:

5. Do not use LRGB combination - it washes out the colour - CieLab combination with masks is the way to go.

Thanks, I had no clue.

44 minutes ago, Barry-Wilson said:

6. Definitely check your DeConv settings or reduce the iterations as you are creating artefacts (worms) and losing detail.

Will do. 

44 minutes ago, Barry-Wilson said:

6. NR and sharpening after stretching in the non-linear stage.

Maybe my sequence wasn't clear. I stretch colour and lum separately and do MLT on masters separately before combining. Should I wait until I have combined?

 

Other questions:

Should I do decon colour master before combining?

37 minutes ago, Barry-Wilson said:

orry of my reply above seems a little brusque

Not at all. Thanks!

[Barry-Wilson]

22 minutes ago, Datalord said:

The benefit of saving the crop and doing it for all at once is that I can reuse masks and DBE processes across. also 

What masks have you created before you might consider cropping the RGB?

The first thing I do with my separate R, G and B windows is to combine them.  Thereafter crop (and apply saved dynamic crop instance to Lum or NB channels as well) and onwards to DBE.

You can then create any mask on either RGB or Lum and you do not need to crop them as the image windows are already cropped.  A step saved and therefore possible sources of error.

24 minutes ago, Datalord said:

I'll try, but my experience is that the gradient will be slightly different between the colours and only by individually running DBE can I get a clean colour master.

DBE will model the gradients across the combined RGB very efficiently if you chose good background samples.

26 minutes ago, Datalord said:

Maybe my sequence wasn't clear. I stretch colour and lum separately and do MLT on masters separately before combining. Should I wait until I have combined?

It can sometimes be an advantage to carrying out NR on the separate channels before combining, it really depends how noisy the data is.  I would try carrying out NR after your combination as an experiment so you are preserving the maximum detail in the lum: with the number of subs you have I would have thought your data was relatively smooth.

35 minutes ago, Datalord said:

Should I do decon colour master before combining?

You only need to deconvolve the luminance channel.

Good luck!

[Datalord]

Well, I definitely see a lot of points that are better in your process. I still need to get a feeling for how to get the best out of it. Here's my first attempt.

Ironically, I think the drizzle somehow manages to give better nebulosity with less noise, even though it theoretically should be worse. Or maybe I should have done more about NR before combining.

[Barry-Wilson]

Well done Thomas - definitely better detail within the galaxy, especially the core.  Better colour balance overall too.

I am sure your image could withstand some small tweaks in curves, "C" and "S", with a suitable mask to protect the stars and background.  Maybe a touch of Exponential Transformation making sure lightness mask is ticked, say at 0.2 or 0.3.

[Datalord]

4 hours ago, Barry-Wilson said:

Well done Thomas - definitely better detail within the galaxy, especially the core.  Better colour balance overall too.

Thank you and thanks for your help!

I agree that this process definitely improves those two parts, but I'm not sure I like this second one better. I'll have to tinker around with when I do NR to get the best of it.