With the help of AI, researchers from Duke University have managed to create a super high-res image from a blurred photo. This was made possible by using an artificial intelligence algorithm that can imagine realistic faces from blurry and unrecognizable images.
“This is the first time a machine developed the ability to recreate such fine details,” said one of lead scientists from Duke’s team.
Named PULSE (Photo Upsampling via Latent Space Exploration), it is capable of creating a resolution 64 times better than the original.
Although it isn’t yet advanced enough to identify people from blurred security camera footage, it is knowledgeable enough to recognise imagery fed to it from satellites, medical scans, and microscopy.
This means those noisy images of distant objects can now be turned into more meaningful pictures. PULSE uses a high-resolution image and then downgrades it to understand and learn rather than doing it the other way round.