[ad_1]
Samsung is in a bit of a controversy over the legitimacy of the Galaxy S23 Ultra’s Space Zoom photos that serve up detailed photos of the moon.
A Reddit (opens in new tab) thread has exploded with a post claiming that the moon shots delivered by the Galaxy S23 Ultra are fake, with the phone applying non-existent details to photos of the moon. Samsung has claimed that it does not overlay images into photos or add in texture effects.
@tomsguide (opens in new tab)
♬ PRETTY BOY FLOW – STURDYYOUNGINN (opens in new tab)
While Reddit user ibreakphotos said this was technically true, in practice Samsung was adding detail into moon shots, which they claim is proven through their own tests. The test involved taking an image of the moon from the internet, blurring it on a laptop then taking a zoomed-in photo of it from afar using a Galaxy S23 Uttra in a dark room.
Given a gaussian blur was applied, there was simply no data for the Galaxy S23 Ultra’s image signal processor or computational photography algorithms to process. Yet the photo served up has a lot more detail in it than could be seen in the original image. As a result, the Redditor branded the Galaxy S23 Ultra’s astrophotography moon shots as “fake” given the detail enhancements were seemingly added.
“Samsung is using AI/ML (neural network trained on 100s of images of the moon) to recover/add the texture of the moon on your moon pictures, and while some think that’s your camera’s capability, it’s actually not,” ibreakphotos concluded. “And it’s not sharpening, it’s not adding detail from multiple frames because in this experiment, all the frames contain the same amount of detail. None of the frames have the craters etc. because they’re intentionally blurred, yet the camera somehow miraculously knows that they are there.”
What happened when we tried the fake moon test
When we tried this on our Galaxy S23 Ultra, we didn’t get the same results. Rather our images of the moon turned out to be a fuzzy mess.
We are just theorizing here, but the differences between out shots and ibreakphotos test results could be down to different software versions of One UI and the Galaxy S23 Ultra camera app. Unfortunately, ibreakphotos doesn’t detail what versions of the respctive software they are on.
It’s also possible that Samsung has turned off the post-photo processing in reaction to this controversy. We have reached out to Samsung for comment on this Space Zoom issue.
Embrace or rage against the machine learning
Now some might claim this is false advertising, and that the Galaxy S23 Ultra isn’t the incredible zooming camera phone you might have expected. But others might shrug and understand that a small camera on the rear of a svelte phone isn’t going to be able to capture shots that even capable DSLRs would struggle with, unless it had some software help.
Given AIs are trained on data, and in this case images of the moon, the processing done by the Galaxy S23 Ultra’s space-related model is likely figuring out what detail should be added based on images of the moon it’s previously been fed.
The result might not be a true representation of what the Galaxy S23 Ultra’s camera sensors can capture, but we’d argue it’s not truly fake either. Rather the moon shots are enhanced with details the camera can’t see or detect. In ibreakphotos’ test case, the details weren’t there but the Galaxy S23 Ultra’s processing is doing the equivalent of filling in the gaps based on what it knows could/should be in certain spots of the moon.
As such, the Galaxy S23 Ultra and its AI might be adding in detail and artificially enhancing moon photos but it’s not simply faking it with a different image. Think of it as someone applying a few edits to Instagram photos to pick out details or deliver more pleasing shots. These enhacements may not be true-to-life, but they aren’t simply fake Photoshopped images either. Whether you think such enhancements are acceptable or duplicitous is down to your tastes.
Again, Samsung has yet to comment on this, but a spokesperson told us that the company is looking into a response.
More from Tom’s Guide
[ad_2]