Google’s RAISR Upscales Images Using Machine Learning

Back in the day when displays would max out at 800×600 resolutions, having images measure 640×480 wasn’t a big deal. In fact it was considered to be pretty big back then. Fast forward to today with monitors reaching up to 5K, clearly an image with a 640×480 would feel almost grainy.

However upscaling images isn’t just about resizing it by hand since that means that it could end up looking very pixelated, but it seems that Google might have figured out a way to do so seamlessly through a technique that they are calling, RAISR, which stands for Rapid and Accurate Image Super-Resolution, which helps to upscale images through machine learning.

According to Google, existing methods of upscaling are fast but not particularly effective because the vivid details in the higher resolution image are lost, meaning that we end up with a blurry-looking photo. With RAISR, Google will use machine learning to identify filters that can help recreate each pixel in high resolution that brings it as close as it can to the original.

This means that the end result could be an image that looks as close to the real thing as possible, but in a higher resolution. So what does Google hope to accomplish with this method? For starters restoring images taken by low-resolution cameras, as well as possibly improving on the pinch-to-zoom gestures used on mobile devices to zoom in and out of images, websites, and more.

Tyler Lee
Categories: General
Tags: AI (Artificial Intelligence), Google