Comment by Modified3019
3 years ago
At the very low quality settings, it's kinda remarkable how jpeg manages to to keep a sharper approximation of detail that preserves the holistic quality of the image better in spite of the obvious artifacts making it look like a mess of cubism when examined close. It's basically converting the image into some kind of abstract art style.
Whereas jxl and avif just become blurry.
It is because JPEG is given 0.5 bits per pixel, where JPEG XL and AVIF are given around 0.22 and 0.2.
These images attempt to be at equal level of distortion, not at equal compression.
Bpps are reported beside the images.
In practice, use of quality 65 is rare in the internet and only used at the lowest quality tier sites. Quality 75 seems to be usual poor quality and quality 85 the average. I use quality 94 yuv444 or better when I need to compress.
You refer to this? https://res.cloudinary.com/jon/qp-low.png
Bitrates are in the left column, jpg low quality is the same size as jxl/avif med-low quality (0.4bpp), so you should compare the bottom left picture to the top mid and right pictures.
Well, that's because JPEG is still using about twice as many bits per pixel, making the output size significantly larger.
Don't get swept away by false comparisons, JXL and AVIF look significantly better if you give them twice as much filesize to work with as well.
JPEG bitrates are higher, so all it means is that SSIMULACRA2 is the wrong metric for this test. It seems that SSIMULACRA2 heavily penalizes blocking artifacts but doesn't much care about blur. I agree that the JPEG versions look better at the same SSIMULACRA2 score.
Humans generally tend to prefer smoothing over visible blocking artifacts. This is especially true when a direct comparison to the original image is not possible. Of course different humans have different tastes, and some do prefer blocking over blur. SSIMULACRA2 is based on the aggregated opinions of many thousands of people. It does care more about blur than metrics like PSNR, but maybe not as much as you do.
Ideally one would use human ratings.
The author of the blog post did exactly that in a previous blog post:
https://cloudinary.com/labs/cid22/plots
Human ratings are expensive and clumsy so people often use computed aka objective metrics, too.
The best OSS metrics today are butteraugli, dssim and simulacra. The author is using one of them. None of the codecs was optimized for that metrics except jpegli partially.
Yes, that was my takeaway from this that JPEG keeps edge sharpness really well (e.g. the eyelashes) while the jxl and avif smooth all detail out of the image.
No, JXL and AVIF keep keep the same level of edge sharpness but without all the blocking artifacts when given the same amount of bits per pixel as the lowest-quality jpeg.