Ever searching for fresh applications for its machine-learning capabilities, Google has turned its attention to what was once the preserve of science fiction fancy – intelligent upsampling, the process of creating a larger, high quality version of a low-resolution image.
According to the tech giant, their RAISR (Rapid and Accurate Image Super-Resolution) technique incorporates machine learning to produce image upscaling results that are “comparable to or better than the currently available super-resolution methods”. Not only that, but they claim that RAISR does so between 10 and 100 times faster – meaning that it can be run on a mobile device in real-time. Impressively, they also claim that the system can therefore reduce bandwidth use by up to 75% per image upscaled – saving significantly on image hosting costs.
Though only rolled out on the Android version of their Google+ social media platform so far, Google have announced plans to expand RAISR across a broader spectrum of services over the coming months.