Hi Jeff, Jon, and JG:
Thank you very much for your kind appreciation, and my regrets on taking so long to reply.
JG, the approach that I used to circularize and erode the stars is due to Carlos Milovic, as described in the following posts:
http://pixinsight.com/forum/index.php?topic=500.15http://pixinsight.com/forum/index.php?topic=2975.msg20394#msg20394I needed to circularize the stars in this image because many of the frames were taken on very gusty nights, which played havoc with guiding along the RA. This resulted in stars with a noticeably elliptical shape that was closely aligned with one axis of the frame.
To correct for this, I tried to follow Carlos' recommendations as detailed in the above posts. A major ingredient is to use the MorphologicalTransformation (MT) with Operator=Morphological Selection, and with two or more "ways" for the Structuring Element, where these are designed to achieve the desired effect. In my case, I used two ways: a line of white squares along the direction of the long axis of the ellipse, to make the stars rounder, followed a square to reduce the overall size of the stars (the small stars were the major problem, so I used 3x3 elements, along with an appropriate star mask). Carlos recommends to use a small Amount with multiple Iterations, and to use a mild Selection of around 0.2 to achieve a gradual erosion. I used Amount=0.05, Iterations=5, and Selection=0.2, and I applied the whole MT operation 5 times. I arrived at these settings after a lot of experimentation.
Rogelio Bernal Andreo has an excellent description of all the parameters in the MT in his "Officially Unofficial" PixInsight reference guide:
http://blog.deepskycolors.com/PixInsight/Morphology.html#MorphologicalTransformationI started the processing with a Deconvolution that had a moderately aggressive aspect ratio of 0.75, with a slight rotation to align with the long axis of the ellipse. I might have been a bit unorthodox here, as I first tried to "objectively" measure the aspect ratio using the DynamicPSF tool, but I found that I needed a rather smaller aspect ratio to achieve a good result. I applied the MorphologicalTransformation in the nonlinear phase of the processing, in fact near the very end of the whole workflow.
I hope this is helpful.
Cheers,
Howard.
