Images are the most used content on the web and so they usually affect the loading time of various websites, in order to get past that Google recent came up with their WebP technology to deliver much smaller images.
Today they have announced an update to that with their new technology called RAISR ( Rapid and Accurate Image Super-Resolution), which uses machine learning in order to deliver sharper low-res images.
Google says that this will result in much better results than other image optimisation methods while being 10-100 faster at rendering images.
To do this, RAISR uses machine learning technology yo train on pairs of images, one of which is a low quality and the other is high, it will then find filters that can be selectively applied to each pixel of the low-res image. These filters come from its database of more than 10,000 low- and high-res images, which it uses to find the best filters to use for other images.
Google says that this will result in delivering low-quality images that are comparable in quality to the original high-res image.
Interestingly, Google are already using this technology on the Google+ website where they are already applying the tech to over 1 billion images this week to reduce loading speeds.
As part of this, instead of viewing a larger image file Google+ will request an image at just a quarter of the size, which it will use RAISR with to enhance the quality.
You can read more about it at the source link below.