Enthusiast reveals the secret of perfect moon shots with Samsung flagships

A few years ago, Samsung introduced the flagship smartphone Galaxy S20 Ultra, one of the features of which was the widespread Space Zoom function, which allows you to take high-quality pictures of the moon with a 100x zoom. Now it has become known that the photos taken with this tool cannot be considered absolutely reliable, since the camera’s AI algorithms add details to the moon images during their creation in order to improve the quality of the photo.

Image source: Reddit

This is the conclusion reached by one of the community members. Reddit, which proved his assumption experimentally. First, he downloaded a high-resolution image of the moon from the internet, resized it to 170×170 pixels, and applied a Gaussian blur. As a result, all clear and small details disappeared from the picture. Next, the enthusiast enlarged the resulting blurred image to the full screen of his monitor, turned off the lights in the room, and took a picture of the monitor. Comparing the original image with the one captured by the smartphone, it becomes clear that the smartphone uses AI algorithms to paint craters and other elements to improve the quality of the photo.

“I hope you understand that in the image above, Samsung uses an AI model to paint craters and other details in out-of-focus areas.”, — said in the message of the user Reddit. He also recalled the existence of a difference in the work of camera algorithms. It’s one thing when, say, an algorithm combines multiple frames to recover detail that would otherwise be lost. The situation is very different when an AI model trained on numerous shots of the moon adds craters and other details to the photo.

In the comments, other users recalled that questions arose about the methods used by Samsung flagships to obtain moon images. two years agoand the reverse output instead whole investigation. And before that, a similar study was devoted to the work of cameras in Huawei smartphones. However, in the age of computer photography, it is hardly worth hoping for the complete “honesty” of images.