Google has released an image compression technology (RAISR) to allow saving without compromising their quality.
Google has recently released an image compression technology called RAISR (Rapid and Accurate Super Image Resolution in Italian Super Accurate and Fast Image Resolution) designed to allow data to be saved without compromising their quality, a novelty tech we already heard about last November.
Image compression of 75%
It is said that the system is able to use 75% of broadband, thus a compression of 75% of the file size, without affecting the quality. RAISR is able to analyze the high and low resolution versions of the images and once analyzed – learns what makes the high quality ones superior, simulating the differences with the low resolution ones. In essence, it uses machine learning to create filters similar to those used by Instagram to make low-resolution images appear as beautiful to the human eye as high-quality and perfectly detailed ones.
The compression method that adapts to images
This new technology works in a similar way to sampling methods that add additional pixels to low-resolution images to compensate for the lack of detail. But, while the traditional sampling system has fixed rules, the RAISR adapts its method on the basis of each individual image, solving the blur problem.
Currently only for Google and Google+
Unfortunately this technology only works with Google+ and Google has stated that it will scale at least a billion images every week. For those unwilling to start using Google+ there will be some more time to wait to enjoy this technology; over the next few months, in fact, Google plans to expand the service algorithm to others of its apps, and we all hope it means adoption on Google Photo.
Also for Twitter image compression with artificial intelligence
But Google is not the only one to use this type of image compression. During the month of June 2016, Twitter bought an English AI startup called Magic Pony which uses similar technology to improve the resolution of low quality videos.
Images are not lacking for training
There is such an abundance of images to allow a mammoth training for machine learning algorithms and the resulting technology offers consumers a simple and tangible benefit in the form of reduced data consumption.