Denoise R G and B individually, or after combining

Benoit Saintot

Hello

On a recent processing of M81/82 https://astrob.in/tu9n8y/D/, I ran TGV/MLT denoise on each linear channel R G B and L before combining, and it worked pretty well.

With the same sequence on M101, with same gear, same exposures, day after, I tried the same way and it gives me weird results when denoising R G and B individually.

So I have a simple question : in which situation is it a better choice to denoise color channels individually before combining RGB, and in which situation is it better to combine before reducing noise on the rgb image ? In general, what is the best moment to apply NR, before ou after combining R G and B?

Thanks for your help!

Benoît

BlockHead

When you denoise a single  color image- you will be applying the same settings to each color layer if I understand. Whereas if you do individually, it is possible to have different settings on each- and I can see how this would give odd results. For example if you did the following... if you have individual R,G, and B images..and you just removed (or used a negative bias) on the 2 or 3 wavelet layer for the B channel.. this will actually change the color balance of the image! (among other things). This is because the relationship between the colors is on a pixel-to-pixel basis between each color channel. If you fiddle with one channel differently than the other...I think results could sometimes not be great.

MureDenoise does not interpolate signficantly between pixels... so this is a little different. All other methods "blur" pixels...so I think you need to apply the same settings to each channel if you do not want them to change. 

So..this is my initial take on this question. It think you can apply DBE differently between color channels in the linear state... and still end up OK. But I think it trickier to do this with significant noise reduction.

-the Blockhead