Night photography is one of the highlights in the marketing of the Galaxy S23 Ultra and the entire S23 series. In recent years, all smartphone manufacturers, including Samsung, have made a concerted effort to improve low-light photography. To prove that it is progressing towards this goal, Samsung uses the Moon photos that it says were taken with a Galaxy S series devices.
However, it seems that the company has not been very honest about it from the very beginning. It is claimed that the moon photos look too good to be true because they are not real anyway. Samsung may be using artificial intelligence (AI) and machine learning (ML) to detect the Moon in shots and then overlay it on top of the image you take. As a result of this, an unreal but spectacular-looking Moon photo emerges.
Behind these accusations is a Reddit user called ibreakphotos. The user set up a simple experiment that anyone can repeat to show that the Moon photos taken with the Galaxy S are not real. The user’s highly detailed post went viral, with several camera samples he posted to prove his claim.
He started his experiment by downloading a high-resolution photo of the Moon. He then downscaled that photo to 170 x 170 pixels and applied a Gaussian blur filter, removing all details from the image. After these procedures, it should have been absolutely impossible to take a great photo of the Moon when you point a Galaxy S phone at this photo.
The Reddit user opened the photo on his monitor, turned off the lights, and moved to the other side of the room. He captured a perfect Moon photo by zooming the screen with his Samsung phone. Despite the blurriness of the original rendered photo, the Galaxy S23 phone managed to come up with an incredibly detailed image of the Moon.
According to the Redditor’s statement, the left square appeared from the right photo.That’s why the Redditor says the only plausible explanation is that Samsung modified the photo with a clever algorithm. Redditor explains: “I hope you appreciate that in the side-by-side image above, Samsung is leveraging an AI model to place craters and other details where it’s just a blurry mess. And I should stress that there is a difference between combining multiple frames to recover details that would otherwise be lost, versus super-resolution additional processing, and here is a tool to recognize the Moon and paste the Moon texture on it (when there is no detail to be recovered in the first place, as in this experiment). There appears to be a specific AI model trained on a series of Moon images. It’s not the same kind of processing that’s done when you zoom in on something else, these multiple exposures, and different data from each frame account for one thing. Special for this month. In short, Samsung’s moon pictures are fake. Samsung’s marketing is deceptive.”
Emphasizing that the addition of details in such an experiment, where there would be no detail, is entirely thanks to artificial intelligence, rather than an optical process such as multiple frame exposure, Redditor ibreakphotos said that since only one side of the Moon is seen from the Earth, it is necessary to train the model on other Moon images and have a similar view to the Moon. it says it’s very easy to paste this texture once the image is detected.
The Redditor didn’t specify which phone Samsung used to prove the moon photos were fake, but a different Redditor did the same a few years ago.
After that, Samsung will have to wait for a response and the company will have to prove that the Moon shots with the Galaxy S23 Ultra and other high-end phones are real. Otherwise, he may have some serious explanations to make.
If Samsung had made it clear from the start that the Moon shots were artificial, this wouldn’t have been an issue. The Galaxy S23 Ultra has a great camera that doesn’t need any of these tricks, and it doesn’t need to take fake Moon photos that just tarnish Samsung’s reputation.
35695
