Regional Deconvolution

[fneyer]

Hi Juan,

Could you implement a possibility to apply deconvolution based on regionally different PSF models? As I understand, you already estimate hundreds / thousands of psf profiles of stars all over an image during ImageIntegration or other processes. Why not use this information to apply deconvolution based on such a regionally varying psf model?

Background: In a perfect system, all stars can be characterized by a single psf model that is valid throughout the entire image. If there is a small sensor tilt or another (even minor) misalignment of an optical element, the latter is not valid anymore. With regional deconvolution, this could at least partially be accounted for.

It would be a nice addition!

Cheers,
Fabian


PS: Thanks for your continuous work on PixInsight!

[dave_galera]

Fabian,  Have a look at this script I think it will do what you want.

https://pixinsight.com/forum/index.php?topic=12735.0

[fneyer]

No, I'm aware of the scripts by Hartmut. The script you mention creates a 'single' PSF that represents the 'whole' image. My point is that there is not a single PSF that is valid for all areas of an image. For example one corner might have elongated stars due to some optical misalignment. I'm talking about a continues PSF model over all regions of an image (so a model of multiple local models) that could be used by deconvolution.

Fabian

[tommy_nawratil]

super idea, Fabian!

Analysing the PSF's of stars throughout the picture and creating a PSF model similar to the background model in DBE
would be an incredible help for processing pictures that suffer of misalignment/distortion.
I remember that Newtonians I rented in Namibia which were unstable, it took me ages to de-distort and restore stars
at the frame edges in postprocessing. 
However, it might be an interesting mathematical task to realize such an multidimensional algorithm. 

Tommy