News for hackers. Stuff that matters.

A comment on Samsung “space zoom” moon shots are fake, and here is the proof. Parent comment

Re:Samsung “space zoom” moon shots are fake, and here is the... by Cyykratahk on Saturday March 11, 2023 @09:09AM (#35108391)
But the AI should not have learned to apply a Gaussian deconvolution kernel. If anything it should be applying a lens-based bokeh kernel instead. A true lens blur does not behave like a Gaussian blur.
[ Parent | on HN ]
Re:Samsung “space zoom” moon shots are fake, and here is the... by gus_massa on Saturday March 11, 2023 @09:23AM (#35108483)
They don't get an exact reconstruction of the original image. What happens if you apply Gaussian blur and then try to undo it with a bokeh kernel?
[ Parent | on HN ]
Re:Samsung “space zoom” moon shots are fake, and here is the... by seba_dos1 on Saturday March 11, 2023 @09:52AM (#35108699)
A mess.
[ Parent | on HN ]
Re:Samsung “space zoom” moon shots are fake, and here is the... by gus_massa on Saturday March 11, 2023 @10:57AM (#35109374)
A mess like in the OP https://imgur.com/ULVX933 or something worse?

(Just in case, original image https://imgur.com/PIAjVKp )

[ Parent | on HN ]
Re:Samsung “space zoom” moon shots are fake, and here is the... by seba_dos1 on Saturday March 11, 2023 @12:32PM (#35110546)
You certainly wouldn't be able to recover any detail this way (in fact, deconvoluting Gauss after taking a photo of the picture displayed on a computer screen won't give you such good results either - try it yourself).
[ Parent | on HN ]