Hi Sander,
Now I have a new user, a terrific bug hunter, and a C++/JS developer, all in a single operation, so what else can I ask for?
Thank you for your support, and welcome aboard.
There's nothing wrong in the way you're using ACDNR. It's just that you have defined a mask that is too restrictive. I have reproduced your parameters and I get just the same poor noise reduction.
Try with a more permissive mask; for example:
http://forum-images.pixinsight.com/legacy/acdnr-sander/01.jpgwhich gives you a reasonable noise reduction:
http://forum-images.pixinsight.com/legacy/acdnr-sander/02.jpgHowever, for this image you actually don't need a mask:
http://forum-images.pixinsight.com/legacy/acdnr-sander/03.jpgAs you see, ACDNR performs very well without the help of a mask for images like this one. A mask is required when there are more differences between high-SNR and low-SNR regions; for example, when there are subtle variations over bright areas with little noise. It's a matter of practice.
Note that I've used edge protection overdrive parameters on both sides. This has helped in protecting star profiles. I've used overdrive also in the chrominance, to protect small stars from becoming grayed out:
http://forum-images.pixinsight.com/legacy/acdnr-sander/04.jpgThe overdrive parameters exaggerate edge protection. With overdrive=0, the algorithm tries to achieve the same contrast as in the original image for protected edges. When overdrive > 0, contrast is increased for protected edges, which has a sort of sharpening effect. Overdrive can help, but these parameters must be used with great care. If in doubt, better don't use them.
Hope this helps.