As the title would suggest, Google has released a new open source JPEG encoder called Guetzli that uses a new open source algorithm to create JPEG files that are “35% smaller than currently available methods” (though if you dig a bit you’ll find they actually say it’s more typically 20-30%). That’s according to Google’s own literature from their Research Blog, and that’s a significant claim if for no other reason than there exits JPEGmini.
The desire to have higher quality JPEG files without the added pounds of large file sizes is a fairly obvious one, especially for photographers; smaller files make for more space and particularly it means faster loading times for images on webpages. In a world where attention spans are as fleeting as a lightning bolt, this matters, and is one of the reasons I run JPEGmini for all my image exports for web images.
For those who are not familiar with JPEGmini, it is, at its heart, a program that compresses your JPEG files to the utmost point before losing any perceivable quality. The critical word there being “perceivable”. That’s all it does, and it does it without much fanfare. Like a worker ant, it just falls into line with your workflow and gets on with the job with a manner of efficiency not typically found outside a Japanese car factory. If it were a Japanese worker, it would surely be prone to Karōshi (death by overwork).
Now, this isn’t meant to be an advert for JPEGmini, but before photographers get too excited by this new Google offering I think it’s only prudent to mention this because it would seem for our purposes JPEGmini is Google’s real competition here, and it’s hard to imagine how much better, if at all, Guetzli is than JPEGmini – which is also using quality measure algorithm guides and drives the standard compression to the point that eliminates all the redundancies beyond human vision. That’s essentially what Google is saying Guetzli does:
The visual quality of JPEG images is directly correlated to its multi-stage compression process: color space transform, discrete cosine transform, and quantization. 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. However, while Guetzli creates smaller image file sizes, the tradeoff is that these search algorithms take significantly longer to create compressed images than currently available methods.
When I was learning about JPEG mini I spoke to Eli Lubitch, a previous VP at Kodak and R&D at Scitex, an image scientist, and now President of BEAMR which makes JPEGmini, and this is what he had to say about it,
We will stop short only just before there are any noticeable visual artifacts to humans, because that’s what the quality measure is aligned with. Other compression technologies in programs such as Lightroom and Photoshop will ask for a percentage or some quality parameter. With JPEGmini they don’t ask any questions, so sometimes they’ll save 90% and other times it’ll be 70% or 60%, because the algorithm commits only to the quality.
Committing only to quality seems to be just what we photographers want more than anything, and the fact JPEGmini makes it easy and fast is also a boon. Actually, the ease of use is quite a big deal when you’re time-pressed. JPEGmini has the option for an integrated Lightroom plug-in and one for Photoshop in tandem with the stand-alone application, and it processes all the images really quickly. Given that fact it’s worth noting that in order to get Guetzli it’s suggested you download and install Homebrew, and then you need to build the Guetzli binary.
And then in addition there’s a warning that “Guetzli uses a large amount of memory. You should provide 300MB of memory per 1MPix of the input image.”
But it’s free, and that’s great, and where it clearly has JPEGmini beat, but how much is your time worth?
My personal opinion is that it’s always great to push these boundaries and develop more, but right now I can’t imagine it being much better than JPEGmini for the average photographer.