After many hours going over the various steps I started to figure a few things out in Pixinsight while reprocessing the Helix data I have collected (94 X 300s of 5nm Ha and 41 X 300s subs of 3nm O3). First it seemed I was just adding noise to noise and I could have done a lot better to limit it at each step. With that in mind I worked out better how to use MLT (noise reduction targeting the smaller layers) and also didn't overstretch the image (making adjustments in with the Screen Transfer Function before applying these settings to a histogram stretch). I also made a lot more use of masks to bring out detail and suppress noise. The HOO version of this object also seems rather more satisfying. In the end the simplest approach seemed to work the best. I wonder if that matches others experience with processing this object in particular? https://www.astrobin.com/full/p6idc2/B/
Thanks Chris. I was kind of surprised when I checked the image against some of the deeper Ha images. Not super clear or anything but something coming through :-)
Thanks Daniel. Yeah, masking detail while reducing background noise is useful as is masking background (not always just an inverted mask) to sharpen detail. The biggest thing for me was just not to stretch it too much. The Screen Transfer function’s automatic stretch was a bit much for this object as outer details are very close to background nosie, so I manually adjust the stretch to suit.