Deconvolution on the full image not equal to the preview

[wkrispler]

Hi,

I am trying to do a deconvolution on a RGB image with target "RGB/K components".

I made a preview as usual and the settings worked fine. When I applied the same settings to the full image
I had red borders on the upper edge of the stars.

The process takes very long so it is very cumbersome to try the settings on the full image, tweak them and try it all over again.
I never had the issue on luminance images...

Any idea?

Thank you,

Wolfgang

[Alejandro Tombolini]

Hi Wolfgang,
Yes, Deconvolution works different in a small preview than in the all image. What may help you is to set the parameters in a preview and then jump to another preview, may be double in size, and adjust again the parameters.
In this post you can see an example on what parameter needs to be adjusted depending on the effect that has to be controlled when you apply the deconvolution algorithm.
When you are working in a RGB image and experiment red rings when goes from the preview to the full image, that may be is because Deconvolution is generating a dark ring in blue channel on some stars. This processing example could be useful in that case.
Hope this helps.
Saludos, Alejandro.

[Warhen]

Hey Wolfy! Good advice from Alejandro. Decon is one of a couple of processes I'm aware of that can play this trick sometimes. Another is HDRMultiscaleTransform. And as Alejandro explains, it's due to the statistics provided by the small preview vs. what actually happens on the whole. Try his suggestion on a larger preview. If all else fails, the SubstituteWithPreview script can work wonders, at least for the object of interest, though it would be nice to have all of the stars deconvolved as well. 

[wkrispler]

Thank you Alejandro and Warhen.

What I am also wondering is if on an RGB image it is necessary to choose "RGB" as target as if I choose "Luminance" it also works!

What is the difference?