@halide this is a bad methodology. When claiming they are adding fake moon detail, they need to test that with an image from the backside of the moon or something to see if it overlays structure that isn't there.

This just looks like image sharpening to me.

Edit: oh shit, they did test it with a cut-off mon and a full moon in the same picture and the phone only did detail recovery on the full moon, so samsung is doing more than just detail recovery.

@Mattmaber so what?

Digital stabilization is Cheating?
Noise-Reduction is cheating?
Color Grading Photos is cheating?
Color Dodging in darkrooms is cheating?
Using a tilt shift lens is cheating?
Seasoning food is cheating?

You're digitizing analog photos. Why don't you just send me a copy so I can look at the analog thing? You're cheating!

Yes, phone cameras are bad. Everyone knows this. Samsung, Apple and Google would be stupid to not use the compute on their devices to improve images.

@Mattmaber okay you kind of struck my nerve with cheating there, as this is an argument a lot of people tend to make.

I think capturing an image that is good enough of the moon to be able to figure out its orientation anddo the detail recovery from a phone camera is still pretty wild and a huge achievement. You would not be able to do that on what apple is capturing on their sensor.

But the sensor data itself is not as good as the result suggests and they are helping out with training data.