Guetzli vs. MozJPEG | Hacker News Reader

Guetzli vs. MozJPEG

pixelz.com

32 pointsluu19 comments

The benchmark doesn't make sense at all. The encoders under the test were set to the same JPEG quality numbers, but this is a completely meaningless metric, and JPEG quantization matrices were not even examined to ensure that the quality is objectively comparable. See my blog post from 2008 regarding this very issue: https://patrakov.blogspot.com/2008/12/jpeg-quality-is-meanin...

(2017)

This doesn't seem like a very good benchmark. First, they appear to be recompressing jpgs rather than starting from the original input image, which seems incompatible with the idea of a compressor that spends the time budget where it does the most good. The places where the high quality is needed the most have already been compromised by the original jpg compression.

Second, they're only considering the output size but not the output quality. Sure, they're all using "quality 90". But why do we think that "quality 90" means the same visual quality across all these compressors? (And what visual quality metric would it be constant for?)

You know. I was actually quite annoyed ( to say the least ) with the post. For one the post is lacking a date, and you have to search yourself it was published in April 2017. And without a date the article is completely lacking context because Guetzli [1] hasn't been worked on for 5 years. And as [2] mentioned its work and derivative was ultimately merged into JPEG XL, which is a very decent image format. ( People should definitely check out JPEG XL if it is not on your radar yet )

But then I notice it was Dan luu who submitted it, which likely means there must be something much deeper than is what is shown on the surface. So what is the context here ?

[1] https://github.com/google/guetzli

[2] https://news.ycombinator.com/item?id=30622303

FYI: Guetzli seems to have not been touched in ~5 years.

See https://github.com/google/guetzli/tree/master/guetzli

It evolved into Butteraugli which was incorporated into libjxl, which powers the reference JPEG-XL encoder. It also optionally powers libaom's encoding, which is in turn used by libavif and libwebp2 (for the WIP WebP2 image format).

Fuck that site. It has autoplaying sound, and a popup that you can't close.

Also a custom scrolling implementation that scrolls a longer distance than normal, and forces a scrolling animation, absolutely horrible.

Guetzli means cookie in Swiss German, I wonder what inspired the devs to choose that name

Other compression algo is Brotli (or brötli).

Also popular Swiss pastry.

Brotli, Zopfli, Guetzli, Butteraugli, ... Google's Zurich branch uses this particular naming scheme for their compression research projects.

What other swiss foods ending in *li can we expect them to name future compression algorithms?

I'm guessing Gipfeli

It exists! https://github.com/google/gipfeli

It was aiming for better compression than LZ4 at much higher speeds than deflate. Now brotli and zstd on their faster settings mostly beat it in that niche.

Wow haha, that is amazing. Thanks for the find!

Weggli(Small bread tasting a bit like a Zopf), Hörnli(The kind of pasta americans use to make Mac n Cheese), Mütschli(Small bread which is a bit harder than a Weggli), Schnitteli(Piece of bread with jam or honey etc.), Schöggeli(Small chocolate), etc...

Basically you can add -li to almost any word in swiss german and there are lots of words already which end with -li. It's a way to make a word sound "cute".

As an example, "Auto" means car, but if we call it "Autöli" it means that it's either a toy car or just a little(,cute) car :)

To me the images produced by Guetzli feel "off", like I'm being tricked into perceiving a higher quality. Maybe this is what is meant by "perceptual psycho-visual encoding".

offtopic question but can this be used to compress base64 images?

Yes! Decode the base64, run through mozjpeg, re-encode. For PNGs, I've had the best results on average with pngwolf-zopfli: https://github.com/jibsen/pngwolf-zopfli

off topic but has anybody turned a series of compressed base64 into a video stream?

Reply on news.ycombinator.com