Google has begun rolling out a new method of reducing the size of images loaded on mobile devices, implementing a machine-learning upscaling algorithm called RAISR to drop filesizes by up to 75 percent.
Originally unveiled back in November
by Google's research team, RAISR - Rapid and Accurate Image Super-Resolution - is designed to cleverly extrapolate the missing information from a low-resolution image in order to produce a higher-resolution version without any blurriness or blockiness. Equivalent in quality to the best rival super-resolution systems, the company claimed at the time, RAISR runs around 10 to 100 times faster - meaning it's possible to implement it on-the-fly, rather than limiting its use to batch processing of files for later consumption.
Unsurprisingly, that's now exactly what the company has begun doing. In an update to its Google+ social network, Google has implemented a smart image resizing system whereby a client device requests a version of the original image at a quarter of the original resolution then uses RAISR to upscale it back to full resolution again. The process is, naturally, lossy, but as with other lossy compression algorithms designed to drop image file sizes like JPEG the gains are impressive: At little to no loss of discernible fidelity, a 100KB JPEG drops to just 25KB.
'By using RAISR to display some of the large images on Google+, we’ve been able to use up to 75 percent less bandwidth per image we’ve applied it to,
' claimed Google+ product manager John Nack of the new system. 'While we’ve only begun to roll this out for high-resolution images when they appear in the streams of a subset of Android devices, we’re already applying RAISR to more than 1 billion images per week, reducing these users’ total bandwidth by about a third.
Nack has indicated that the RAISR system will be rolled out more broadly in the coming weeks, though it is not yet clear whether the company plans to implement it in its Chrome browser software for use outside the Google+ ecosystem.