Google Brain engineers have developed a system based on neural networks which can significantly improve image resolution and zoom in tiny photos preserving their details.
Google system works in two phases:
- one algorithm matches the image with high-quality images uploaded in the Internet to approximately define its final look;
- the other algorithm is trying to precisely model the final image by means of the received information.
As a result, these unique algorithms can help to essentially improve the quality of a pixel image. Although, image modeling results are far from being perfect, in some cases the system identifies and depicts people’s faces quite accurately while editing enlarged 8×8 image.
Google experts are likely to develop their product in the future which means that we will soon get a wonderful instrument for edition images and recognizing faces.