Shulou(Shulou.com)11/24 Report--

CTOnews.com, March 23 (Xinhua) earlier, a photo of the moon taken by Samsung Galaxy S23 Ultra phone was suspected of being falsified, although officials denied this claim, but some netizens still did not buy it.

Eric, a Youtube blogger and engineer with 12 years of hardware engineering experience, recently used mathematics and logic to analyze how to distinguish between real and fake pictures in the video, and pointed out that Samsung may have been unfairly criticized.

CTOnews.com noted that before starting to explain, Eric stated that he was not trying to defend Samsung, but to give people a better understanding of why the images showed such an effect. First of all, he asked the audience how to distinguish between real and false pictures. Then he showed a picture of the moon taken with iPhone 14 Pro Max, with only one white spot visible.

However, although the photo taken by iPhone 14 Pro Max is real, it does not accurately reflect what the moon looks like when people take it, and that is what the camera is for: to preserve and revisit the memories you saw with your own eyes.

The tubing blogger also mentioned that Samsung explained the technology it used in an article. In order to enhance the effect of moon photos, the camera took at least 10 pictures and used image processing technology to remove defects and noise. The clearest parts of these moon photos are then combined into a separate image and further enhanced using AI, which has been trained to identify the various stages of the moon.

Eric also mentioned u / ibreakphotos, who used Gaussian blur and other steps on Reddit to prove that Samsung misled consumers with "enhanced" moon photos. However, Eric said that netizens using Gaussian blur to imitate moon photos is the opposite of what Samsung does: Samsung uses a convolution neural network. Eric also said that when netizens reduced the moon image to 170x170 pixels, they lost a lot of detail and applied Gaussian blur. When dealing with such a small amount of detail, Samsung AI must make the best guess about what it is dealing with, so the end result will not exactly match the original scaled-down image.

As Samsung's Convoluted convolution neural network is running, the final picture shows a brown hue. The reason why the image is brown and the edge quality is declining is that the Samsung convolution neural network is at work. When removing the brown tone, you can see very close to the picture before the netizens applied Gaussian blur.

The problem with this series of events is that Samsung misled consumers into believing that the camera on its flagship smartphone was doing all the work, not the AI. This may make users think that if the moon can be taken so clearly, photos taken with 100x zoom can be so clear, when in fact this is not the case.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.