StarTools

Hello,
I was following the Scott Rosen method to post process M31 data when I encountered an old problem: when attempting to make heavy denoise in the RGB part of the image, things turn to purple

here an example:
there is a trick to avoid this (and I'd like to have your opinion): make a normal denoise, save the image, load them with tracking off and apply heavy denoise and you get what you want
is there a more "canonical" approach? in the forum posts I never read someone with a similar problem, still it happens to me everytime a try to boost smoothness or increase the detail loss sliders.

I made a few tests increasing grain size and the problem became less visible (because you don't need to push the other params that much) but still existing

thoughts?

Michele

It might be the very high "Color Detail Loss" setting you are using - this will make the color information lose accuracy, which can cause incorrect coloring. The "Brightness Detail Loss" is also quite high. You Grain Size parameter is very small and does not suit the image (noise grain can still be easily seen).

In the first screen of Denoise, you are asked to increase the Grain Size parameter until you can no longer see the noise grain. The image will become blurry, but - in that screen - that is desired behaviour. Noise grain carries through into larger image scales (becoming clumps, spots and blotches). The point of the Grain Size parameter is to let Denoise know when you, as a viewer of the image, can no longer detect those clumps, spots and blotches at all. Denoise will keep any larger scale detail larger than that size.

Setting the Grain Size parameter too low will cause Denoise to keep visible grain above that threshold. In your case it will keep grain size over 3 pixels in size, but that grain is still very much visible.

Applying Denoise after tracking is sub-optimal, as Denoise no longer can use the Tracking data mining information to improve the noise reduction. So it is not really a good workaround.

Does this help?

thank you Ivo, I always try to avoid increasing grain size too much because - as you said - the image becomes blurry, but in case of heavy noise reduction for layer module usage I will work on grain size, thanks!

Michele