New Google Technology Makes Blurry Images Clear for Security Impacts

Scientific and technological developments, according to foreign tech website reported February 7 arstechnica, and now Google has developed a new software technology will be able to rely on artificial intelligence learning fuzzy picture becomes clear, for the security industry will have far-reaching significance.

New Google Technology Makes Blurry Images Clear for Security

From the given sample image, Google's software is able to create pixels based on the original image, restoring the original 8*8 pixel size image to amazing detail. We know that if artificial intelligence is unable to create details that the original image lacks even if it is more powerful, how does the artificial intelligence technology make the photos clear?

According to reports, to achieve this effect, two neural networks are required to work together. The first stage involves the Conditioning Network. The system will search for other similar high-resolution pictures that have been compressed to find patterns, The similarity of colors.

The second phase is called "PriorNetwork". This phase will use the details of the high-resolution picture to fill the pixels of the blurred picture.

It is reported that the pictures of the two neural network stages will be further integrated into a picture, which contains details that are reasonably added to achieve maximum resolution reduction.

Google’s testing of this technology in the real world was quite successful. In one test, when human observers were asked to look at a real celebrity high-resolution photo and a computer-recovered photo, 10% of them were puzzled. For the bedroom photos, up to 28% of human observers are confused by computer pictures.

It is worth noting that the computer super-resolution image is not a true original image. The additional details added in the field of image processing are termed “phantoms,” but they are “guessed” by the computer. In other words, the system considers the new image. Clarity is enough to "deceive" human eyes. Although it is not 100% true, it can still play a supporting role in areas such as finding criminal suspects.