A group of scientists have used artificial intelligence to create a new tool that can make blurry photos sharper.

Called EnhanceNet, the software uses texture analysis powered by AI which is able to upscale the quality of the image automatically.

[Read more: Can you spot a fake photo?] 

Researchers from the Max Planck Institute for Intelligent Systems in Germany have now written a paper on the subject, detailing how their system works.

(Mehdi SM Sajjadi, Bernhard Scholkopf, Michael Hirsch)

According to the paper, published to arXiv, EnhanceNet uses “automated texture synthesis” that aims to recreate realistic textures in images, rather than “over-smoothed” photos which the researchers say other techniques produce.

[Read more: Fake photo wins Nikon competition]

The software is also powered by a technique known as adversarial training, which in essence pits the different AI networks within the software against one another, with the competition ultimately helping them learn together.

The researchers say this technique, when applied to images, helps achieve a “significant boost in image quality”.

One of the authors of the study, Mehdi SM Sajjadi, told FastCoDesign that the technology could one day find its way into programs such as Photoshop, and could in theory even be used to unpixelate faces in photos, he claimed.

Read more: Quiz - can you tell real photos from fake ones?