Hi Simon
You really should get into SGBNR and SCNR. If you use them, with wavelets in the sequence SCNR-Wavelets-SGBNR there is almost no kind of noise that you can't remove.
The definition of noise scale, for reduction, is just on wich tool gives you better results (forget SCNR for this). For example, wavelets works better with smaller noise with high amplitude. The key here is that the noise is smaller, or equal size, than the fainter stars. If you go to greater scales, wavelets can't isolate noise from real features, so noise reduction becomes a very difficult task. By the other hand, SGBNR works bets with large noise, wich usually has lower amplitude (contrast) than real features. Becouse SGBNR works with intensity thresholds, if you have small noise with great amplitude it will be considered as a real feature.
Usually, the high frecuency noise lives in the firsts wavelets layers. If you use a dyaic sequence, the 4th layer will barelly show it, so avoid noise reduction for scales larger than it. Of course, you have to inspect the layers for a better cut-off. =)
As conclusion, the difference between small and large noise isn't just the size, but also the amplitude of it. Also it depends on how much can be done with wavelets without bluring the image.
Now the LRGB combine... Use import channels. Create a RGB image and import the chrominance channels as common R, G and B. Then, import the luminance, using the CIE L*a*b* color model (L* channel).
Image registration is supported by a manual tool (DynamicAlignement) in PixInsight Standard. Juan is working on a fully automated tool. A partial result is a little comand line application called "R", made using PixInsight's Class Library for C++. It aligns two or more images using translations, rotations and scale factors, along with local distortions. It works fully automatically, but the images cannot be too much different. We are working in a more general solution.
In the meanwhile, DA will be upgraded, and become a semi-automatic tool.
I'll take a look at your software. Seems interesting. BTW, do you know C++? The next phase of PixInsight's development will be more open to colaborations from the users, making theyr own modules, so you may incorporate astrowave's funcionality here (and much more, of course).
