Mobile phones proved long ago that they are an amazing tool to perform the function of a camera. The sensors and processors that mount the high-end models continue to surprise users with high-quality and versatile photographs, and the advent of artificial intelligence is allowing extremely surprising results to be achieved. But, Has artificial intelligence gone too far when applying touch-ups?
The Moon and the excess of Photoshop
In recent weeks, several cases of users who are dissatisfied with the operation of their phones have been appearing. But curiously, not because they did it wrong, quite the opposite: the photographs they obtained were too good to be real.
On the one hand, we have the case of Alvin, a Twitter user who, after taking photos of the moon with his LIVE X80 Prodecided to make a comparison with the Samsung Galaxy S23 Ultra to see how far his suspicions were true. And it is that, for Alvin, the VIVO Supermoon mode did something strange to the photo. Was the phone enhancing the image of the moon in any way?
In the video below, you can see that I made an overlay of both pictures.
I set the fixed transparency of the Galaxy S23 Ultra’s picture to 60% and I put the vivo X80 Pro’s picture above it.
You can see how different both pictures are when I played with the transparency slider… https://t.co/I63egvjXIz pic.twitter.com/if0cgElQ0A
—Alvin (@sondesix) March 9, 2023
After comparing the photo obtained from the Galaxy S23 Ultra and the one from the VIVO X80 Pro, he discovered that the photos presented small differences, since it showed that there were areas that appeared darker and with different craters. Was ALIVE manipulating the image too much?
But Samsung is not saved either
The evidence was pretty self-explanatory, but what Alvin didn’t realize is that Samsung also applies pretty aggressive effects to moon photos. That is the conclusion reached by the Reddit user ibreakphotos, who with a series of tests has shown that the space zoom mode completely manipulates the results when taking a photo of the moon. Do you want to know why the moon appears so well in the Galaxy S23 Ultra photos?
The test consisted of the following:
- A real image of the moon has been downloaded in high resolution.
- I scale it down and apply a Gaussian blur filter in Photoshop.
- Increased the size of the new photo to make the moon bigger.
- He photographed the computer screen with the room dark from a Galaxy S20 Ultra.
- In the photo, the blurred Moon appeared perfectly sharp.
What should have been a blurred image of the moon ended up being a perfectly recognizable and more or less clear image of the Moon and its craters. But how was this possible? Basically the phone’s software used artificial intelligence to generate a perfect image of the moon based on a large number of similar images, in order to represent the craters and details of the satellite in the best possible conditions.
Samsung doesn’t explain it so well
The official explanation from the manufacturer for its Space Zoom mode is that the phone captures several frames in just a few seconds and uses artificial intelligence to improve details and reduce both blur and noise in the photo. This would be great if it were literally like that, but ibreakphotos tests show that the results can turn out to be completely generated images from scratch.
This opens up an important debate, since the AI would be applying too many touches that are not available to the user, giving life to unreal images that would not correspond to reality. This is something that has been debated for many years with the appearance of Photoshop, and as always, it is not very clear to what extent calibrating a white balance or placing non-existent elements on the photograph can be considered extreme retouching.
Be that as it may, it seems that behind the striking function of taking photos of the moon there is quite aggressive processing in many manufacturers, and although the results are great, this is something that should be indicated in a more detailed and transparent way for the users. consumers. Don’t you think?