Google’s new algorithm can reduce JPEG image file size by 35 percent


Since most of us in the third world countries have dismal internet speeds, companies like Facebook, Google, and Opera implement new ways to speed up webpage loading times. Opera does server-side webpage compression for fast page load times, while Google and Facebook have implemented AMP and Instant Articles, respectively, which load faster on smartphones. Google thinks that the easiest way to load webpages fast is to compress image size without affecting its quality. It took the first step in that direction in 2014 by introducing WebP image format, which could shrink an image’s file size by 10% without affecting its quality. Now, the company has taken one more step in the same direction.

Google has now showcased Guetzli, an open-source algorithm that encodes JPEGs in such a way that their file sizes are reduced by 35%, and this doesn’t affect the image quality. As Google points our in its blogpost, WebP needs both client and ecosystem to change for the implementation. However, Guetzli is like its Zopfli algorithm (it is used to compress PNG and gzip files), which means images don’t need to be converted into a new format to take the smaller file size advantage. Google also pointed out that Guetzli’s psychovisual model “approximates color perception and visual masking in a more thorough and detailed way than what is achievable” in current compression methods.

The only disadvantage of the Guetzli image compression algorithm is that it takes a little longer to complete the compression, but Google says that humans preferred Guetzli’s results more than libjpeg. You can see the difference between an uncompressed image, libjpeg compressed image, and a Guetzli’s compressed image below.

Uncompressed Image vs. libjpeg vs. Guetzli
Uncompressed Image (Left) vs. libjpeg (Middle) vs. Guetzli (Right)

[Source: Google Research Blog | Via: Engadget]