This was an interesting experiment.
I notice that when I get to the Create Reference Frame stage of Registax and do a 1.5x Mitchell resample, the composite of the best 50 frames for the reference always looks great, but then the finished version (comprising more frames) falls away after I apply any additional processing.
That got me thinking. When it gets to the stack stage, perhaps I should only take the best 50-100 frames if I'm going to do my resampling in Registax.
So here's the result. It's a little grainier (coz of fewer frames) but overall less afflicted by onions rings.
Interesting
