Hi Juan,
thank you for your appreciation. This image would be not this image without your regularized Richardson-Lucy deconvolution. I hope to remember all the details you ask.
For the central dust lane of the galaxy and for the background galaxies I have resized x2, aligned via Nearest Neighbor and summed the best 200 minutes (better FWHM and star shapes).
Then I have resized to the original size the "drizzled" sum with Nearest Neighbor again.
After a simple Histogram stretching I have applied ACDNR filter (with StdDev 0.1 - Amount 0.1 - Iteration 16) repeated 20 times via ProcessContainer + 3 iteration of ACDNR filter again but with StdDev 0.2 - Amount 0.1 - Iteration 16.
Then I have applied the regularized Richardson-Lucy deconvolution to 4 copy of the original image ACDNR filtered with four different StdDev/Shape approaches:
image 1: StdDev = 1.80 - Shape = 1.90 - Iterations = 10
image 2: StdDev = 1.50 - Shape = 1.80 - Iterations = 15
image 3: StdDev = 1.30 - Shape = 1.20 - Iterations = 20
image 4: StdDev = 1.00 - Shape = 1.00 - Iterations = 25
I don't remember exactly the amount of noise thresold and deringing parameters I have applied, but for the noise (wavelet regularization) I have selected 2 wavelet layers with a threshold around 2.5/3.0 for the first wavelet layer.
Then I have mediated the 4 deconvolved images with PixelMath, masking the outer portion of the galaxy and the brighter stars. Finally I have summed the resultant image (deconvolved) with the 270 minutes image histogram stretched and acdnr filtered.
About the new deringing algorithm, I think that it's more effective, even if some parameter are less intuitive, as for "Deringing threshold" or "Stretch". I would appreciate the return of the old slider "Amount". The preview mode seems to be a powerful tool not only for "previews" but also for creating different kind of masks. Hope that you understand my bad English... and thank you! Regards,
Enzo
