JPEG trade-off by Google Thursday, 06 April 2017

News article written by Corbett Communications. The statements made or opinions expressed do not necessarily reflect the views of Engineers Australia.

One of the internet’s earliest digital netizens, the ubiquitous JPEG - first released in 1992 - has been improved upon by Google with an open source algorithm that can create high quality images with file sizes 35% smaller than current methods.

The algorithm, Guetzli, which is ‘cookie’ in Swiss German, was released mid-March to enable webmasters to create webpages that can load faster and use less data. The aim is to apply it to photographic content on image-heavy websites which will make the user experience smoother and reduce load times and bandwidth costs for mobile users, because replacing the JPEG format would be nigh on impossible.

What happens with the new algorithm is it strikes a balance between minimal loss and file size by using a search algorithm that tries to overcome the difference between the psychovisual modelling of the JPEG format and that of Guetzli. Guetzli approximates colour perception and visual masking in a more thorough and detailed way than what is achievable by simpler colour transforms and the discrete cosine transform, Google said.

From a practical viewpoint, Guetzli is similar to its Zopfli algorithm that produces smaller PNG and gzip files without needing to introduce a new format, according to Google. But, it added, that it was different to the techniques used in RNN-based image compression, RAISR and WebP which all need client changes for compression gains at internet scale.

The visual quality in the 25-year-old JPEG is directly correlated to its multi-stage compression process (colour space transform, discrete cosine transform, and quantisation. However, Guetzli specifically targets the quantisation stage in which the more visual quality loss is introduced, the smaller the resulting file. However, in reducing the file size, Google admitted there is a small trade-off.

“While Guetzli creates smaller image file sizes, the trade-off is that these search algorithms take significantly longer to create compressed images than currently available methods,” it stated.

Google found in experiments where compressed image file sizes were kept constant, people preferred the Guetzli images over libjpeg images even when he libjpeg files were the same size or even slightly larger.

“We think this makes the slower compression a worthy trade-off,” Google said, adding it hopes the new “explicitly psychovisual approach in Guetzli” will inspire further image and video compression research. Now, some industry sectors are wondering when the now prehistoric GIF will get a makeover.

Author: Desi Corbett



Image: 20x24 pixel zoomed areas from a picture of a cat’s eye. Uncompressed original on the left. Guetzli (on the right) shows less ringing artefacts than libjpeg (middle) without requiring a larger file size. Source: Google

#JPEG #Guetzli #Zopfli #RAISR #WebP #Google