We’ve all seen in movies that computer experts working for spy agencies can zoom in low res photos and videos. They can make the pixelated images clear with some fictional piece of technology.
While it may seem a bit unrealistic, Google Brain team has came up with some new software that can create detailed images from tiny, pixelated source images.
Super Resolution Images from Tiny Pixelated Source:
In the picture below left column has the pixelated 8×8 source images, centre column shows the images made by Google software and the real images are shown in the right column.
It isn’t a 100% match, but the accuracy of details lifted just from a 8×8 image is quite remarkable. The Brain team came up with this super-resolution technique by combining two neural networks.
The first one tries to map the the pixelated source image against other high resolution images. It reduces the size of other high-res images to 8×8 and tries to make a match.
The second part uses an implementation of PixelCNN to try and add realistic high-resolution details to the 8×8 source image. To get the final image, outputs from both two neural networks are merged together.
Do keep in mind that these computer enhanced images are not real. The added details could help make a better guess and nothing more.
