# Point spread function image deconvolution

I would like to deconvolve an image of Saturn.

• I took an image of Saturn: Stack of 50 frames, the angular resolution of the original frames is 1.6''/pixel and the frames are scaled x4 before stacking.
• I took an image of a star (Altair) using the same optical system and the same camera settings, stack of 50 frames, same resolution as Saturn. The separation between the star and Saturn is about 30°. I think that the shape of the star is mostly caused by the chromatic aberration of the scope.
• I deconvolved saturn using Richardson–Lucy algorithm and the normalized image of the star as the point spread function (PSF) of the system.

However, the result is absurd (cf image). Why doesn't this method work? The image of the star seems pretty big. Do I miss a step to recover the PSF ?

• some thoughts; 1) The star image looks like it might be saturated. It's hard to tell from this plot. Can you show a histogram of the intensity values? i.stack.imgur.com/mh154.png from pastebin.com/X1S7LHx6 2) What is the angular scale of your images (arc seconds per pixel) and are the shapes caused by astronomical seeing or diffraction? What is the separation distance (arc seconds) between Saturn and the star? 3) Is the algorithm commonly used to remove the effects of astronomical seeing? If so, what constraints are involved? Thanks! – uhoh Sep 8 '20 at 23:47
• thank you! 1,2) I edited the question. Indeed, the star seems saturated 3) Richardson–Lucy is commonly used in image processing and astronomy but I don't know under which constrains – Nichola Sep 9 '20 at 2:39
• excellent edit, thank you! – uhoh Sep 9 '20 at 4:26
• A 30-degree difference leads to a complete difference in the atmosphere ( $r_0$ and the like). If your telescope was of sufficient aperture to suffer atmos. aberrations, both the view direction and the difference in time of image captures might affect the results. – Carl Witthoft Sep 9 '20 at 15:41
• But if you are dealing with a saturated reference image (the star), try using the nonsaturated region and interpolate (in 2-D) the 'true' intensity pattern in the central, saturated region. Then use the procedure you described. – Carl Witthoft Sep 9 '20 at 15:43