Noise Reduction in Linear state problems

[stevek]

Folks,
I am really struggling to reduce noise from my KAF8300 sensor camera.  With Sony sensors I do not have this problem.
As you can see in the heavily cropped attached screen shots of M81, I have a black "mottling" pattern of noise that I can't seem minimise in the linear stage.

This is three hours of RGB after DBE and photometric color calibration so I am still early in the workflow.  My next step is always to use MLT on the linear image to reduce the noise.  MLT succeeds in doing this on the linear image only to reveal this black mottling pattern.  Using different MLT amounts and thresholds does not seem to make a difference to this.  The only way I seem to know to be rid of this is to be overly aggressive with the black point when doing the stretch and in doing so this kills some of the IFN in the image to which this clip belongs.

Any ideas please folks on how to try and reduce this black mottling?

Please ignore the over-green which SNCR would sort out.

Warm regards, Steve

[bulrichl]

Jon Rista has made a detailed three-part tutorial "PixInsights Tips: Effective Noise Reduction" https://jonrista.com/the-astrophotographers-guide/pixinsights/effective-noise-reduction-part-1/ which in my view is excellent, highly recommended.

Especially valuable: In part 2 (Linear noise reduction) the sections covering TGVDenoise and MMT (not MLT!) and in part 3 (Non-linear noise reduction) the section covering ACDNR. The tutorial includes the preparation of the utilized masks.

Bernd

[Geoff]

I agree with Bernd. John Rista’s site is excellent.

[STEVE333]

I also use MLT for noise reduction with the linear data. However, when the data is quite noisy (like your data, and, much of mine) I find I need to reduce the Threshold and/or Amount. I notice you have 3 Iterations. I've never used more than 2, and, oftentimes only 1.

The reason for doing less noise reduction is, if I "over reduce" the noise at this stage it always causes "plastic noise" in the finished image.

Noise can't be eliminated, but, only reduced. In my experience, the worse the noise the less I can reduce it. In other words, with low noise data it is easy to do noise reduction while with noisy data it is difficult to do noise reduction without causing "plastic background".

Hope this makes sense.

Steve

[freed]

You could also try MureDenoise if you haven't used Drizzle. My CCD is less noisy but still noisy and the J Rista method indicated above works well although you may need to tweak MMT or even consider two passes.