Paul Monckton , CONTRIBUTOR
I write about photography and related subjects.
Opinions expressed by Forbes Contributors
The common JPEG could be about to get a lot smaller thanks to Google’s new software. This could lead not only to a faster Web, but also to direct savings on storage costs for everyone from hosting services to hobbyist photographers and especially smartphone users on metered connections.
20×24 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. Google Research Blog
20×24 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.
Back in 2014, I wrote about BGP, a new file format purported to deliver equivalent quality to JPEG but with much smaller file sizes. The purposes of BPG is ‘to replace the JPEG image format when quality or file size is an issue’.
Fast forward to 2017 and BGP has obviously failed to achieve its stated purpose, with the humble JPEG still firmly entrenched as the file format of choice for the vast majority of users.
However a new, and free to use, compression technology from Google now hopes to revolutionise the JPEG where BGP has failed. Announced earlier this month, Guetzli is a new open source algorithm which creates JPEG files 35% smaller than typical current methods.
What sets Guetzli apart
The crucial difference between Guetzli and BGP is that the latter requires new code to be written before it can be read. Standard browsers and image software would simply fail to read the files without specific support for the format.
Guetzli, on the other hand, continues to use the established JPEG format. So, all software which can currently read standard JPEGs will also be able to read Guetzli JPEGs without modification.
There’s some crossover, in terms of the final outcome at least, with Google’s RAISR technology, announced at the end of last year, which can blow up small images into much larger versions with significantly higher quality than was previously possible.
Both Geutzli and RAISR can cut down significantly on the required size of image files, albeit in rather different ways. There’s also no reason why the two technologies can’t be used together.
How does it work?
The Guetzli encoder works by increasing the level of compression while creating the JPEGs, leaving the standard decompression algorithms for reading and displaying the images unchanged. The increase in compression comes from a new and more sophisticated model of human colour perception than is used by current JPEG encoders.
This results in higher quality images at a reduced file size, but also comes with a tradeoff in speed. Google’s engineers say that Guetzli is currently significantly slower than a standard JPEG encoder. Users of the current Windows version are reporting conversion times of several minutes for a single large JPEG.
Guetzli is potentially great news for anyone who stores or displays JPEG images as the time taken to download and the space needed to store them is significantly reduced. However, there are other options for those who want to create smaller JPEGs.
One such option is JPEGmini which claims even bigger file size reductions than Guetzli, up to 80%, with no loss in perceived quality. The big difference here though is that JPEGmini is a commercial product with prices ranging from $29 for the basic Home User option, up to $199 per month for those wishing to use it on Web servers and large photo repositories.
At the moment, JPEGmini is still a good option for everyday use as it is a polished end-user product, complete with plugins for Adobe Lightroom and Photoshop. However, a free algorithm such as Guetzli, once the speed issues are addressed, will most definitely pose threat to paid-for products like JPEGmini which hope to charge for exactly the same final result.
Visit the Kairos webiste https://cabinet.kairosplanet.com/register/#111b0e