We usually only see this kind of technology on CSI, where there’s a super low-resolution image and then the picture is somehow able to use some sort of special technology to render the blurry image into a high-resolution picture clear for all of us to see. That kind of tech used to be pure fiction, but now it no longer is thanks to a group of computer scientists at Max Planck Institute of Intelligent Systems in Germany that have come up with a crazy algorithm that’s able to make pixilated images into clear and high-resolution images.
The tool is called EnhanceNet-Pat, which works with an artificial intelligence to essentially create high-definition versions of low-resolution pictures. Though the results aren’t perfect, it is still highly impressive.
“The task of super-resolution has been studied for decades,” Mehdi M.S. Sajjadi, one of the researchers on the project, told Digital Trends. “Before this work, even the state of the art has been producing very blurry images, especially at textured regions. The reason for this is that they asked their neural networks the impossible — to reconstruct the original image with pixel-perfect accuracy. Since this is impossible, the neural networks produce blurry results. We take a different approach [by instead asking] the neural network to produce realistic textures. To do this, the neural network takes a look at the whole image, detects regions, and uses this semantic information to produce realistic textures and sharper images.”
The team of researchers train the algorithm by feeding their neural network with a large data set of images to allow it to beef up its knowledge about different textures. The neural network only got to see downsampled versions of the images and then was tasked with unsampling the pictures. Once the network was done, the researchers compared the image the network made with the original and tweaked the algorithm to correct errors that they saw.
Sajjadi believes that this tool could offer a lot in the future, even helping make old movies into 4K.
“There are a lot of applications for this,” Sajjadi continued. “From upsampling old movies to 4K quality, restoring old family photographs that are too blurry when you want to get a large print, over to more general applications such as improving object detection. [It also] turns out that using our algorithm on images makes it easier for other neural networks to detect objects in images, which has wide applications, from Google image search to detecting pedestrians in self-driving cars.”
This is definitely an exciting technology. Hopefully it proves to be a useful tool in the future.
