Google Just Open Sourced Guetzli, A New JPEG Encoder Which Cuts Image Size By 35% Without Quality Loss

Google’s aim to speed up web page load time received a boost with the recent announcement of Guetzli (a Swiss-German word for cookie) which they say is designed to cut JPEG image files by 35 percent. In a blog post by Google software engineers Robert Obryk and Jyrki Alakuijala, they said Guetzli will reduce file size while maintaining quality and browser compatibility and this is contrast with the Google WebP algorithm that dies the same thing with the major difference being that Guetzli won’t alter file format at all.

WebP offers around a third better image compression than JPEG, which can add up to a lot of bandwidth savings and speed improvements, depending on how image-heavy a page is. YouTube was able to cut down page load times by up to 10 percent when it recently started rolling out WebP video thumbnails. Google has also saved several several terabytes of bandwidth every day since switching images in the Chrome Web Store to WebP, and reduced the site’s average page load time by nearly one-third. And when Google switched to WebP within its Google+ mobile apps, it saved 50 Terabytes of data every day.

The researchers added that “From the practical viewpoint this is very similar to our Zopfli algorithm, which produces smaller PNG and gzip files without needing to introduce a new format, and different than the techniques used in RNN-based image compression, RAISR, and WebP, which all need client changes for compression gains at internet scale.”

Using quantisation JPEG compression technique, Guetzli cuts the number of colours in a typical photo file and by so doing, cuts the size unlike WebP that actually cuts the size of the image directly thereby making it difficult for users of photo editing tools. “Guetzli specifically targets the quantization stage in which the more visual quality loss is introduced, the smaller the resulting file. Guetzli strikes a balance between minimal loss and file size by employing a search algorithm that tries to overcome the difference between the psychovisual modeling of JPEG’s format, and Guetzli’s psychovisual model, which approximates color perception and visual masking in a more thorough and detailed way than what is achievable by simpler color transforms and the discrete cosine transform.”

The algorithm is now open source and now they expect web masters and graphic designers to apply it to their content to make load web pages faster. Google Search being the biggest search tool in the world would naturally want to give users a better experience.