I have used it in the past and usually it adds too much grain to my images. Where as Lucy Richardson brings out the detail and keeps the image relatively smooth.
I will give this last image a go on M.E and see what I get.
At a guess, without seeing your data, try radius 1.5 and 6 turns in ME and see what happens. I start around there and if the image looks too blurry then I lower the radius and try again. If the radius gets too small then it will start to amplify noise and you have to go larger or reduce the number of turns.
Because ME includes some lowpass filtering as part of the algorithm you don't want to go fewer than about 5 turns or you'll lose more in the filtering than you gain in the deconvolution part.
The radius will depend on the colour as well - I normally find red is smaller radius than green, which is smaller than blue - this follows the resolution in each channel with red generally having the best resolution as it's least effected by air.
When you find the right spot for your data then you can try changing the number of turns to optimise that, on really good data I can end up with something like 10 or 12 turns, but on not-so-good data it doesn't normally go past 5 or 6.
This all depends on the amount of noise in the image, which in turn depends on how many frames you stacked and how much noise was in the original frames...
That double-ring effect around the edge usually means the deconvolution radius is too large (or the number of iterations is too large).
I forgot to say that the results you get in AI are directly influenced by the wavelet settings you use in registax in the earlier stage... if the image has too much wavelet sharpening then you won't be able to do much more with deconvolution. It's a tricky balancing act, sometimes I go back to registax and change wavelet settings, then load that into AI and see how it goes with deconvolution, then back to registax for another try etc.
Congrats Lester impressive image
