BPG Image Format | Hacker News Reader

BPG Image Format

bellard.org

67 pointstosh42 comments

Bellard is obviously a technical genius. I have no doubt that BPG is competitive on a technical aspect. Politically though AV1 is looking like it has the best chance of unseating jpg. Near universal buy-in by browsers/OSes/hardware makers, hardware acceleration starting to roll out, and patent free.

Lots of other discussion on this though: https://hn.algolia.com/?q=bpg+bellard

Will it necessarily matter though? In a WASM context format starts to matter less since content in arbitrary codecs can be downloaded and decoded.

There are some concerns:

Non-native support requires non-static pages - even for something that will always appear to be the same, things like noscript will kill the page;

Hardware acceleration will need to be done in general terms - something like WebGPU will help, but it will still be less efficient than hardware acceleration (and that could be a problem for mobile).

Also CORS: you can hotlink images from everywhere, but if you try to actually put those bits in a Buffer (as you'd need with BPG) the browser won't let you unless CORS is properly set up for the specific source-target URLs (affects things like image CDNs and so on).

Chrome and Firefox are already implementing AVIF so using WASM to decode a different, more-patented format seems pointless.

Hardware acceleration, as you mentioned, will never be as efficient.

Besides, do you really want to download and parse a codec for every page?

No, not for every page - it will very much be use case dependent.

However, it does present a viable path to the introduction of new codecs independent of the bogged up process that still has us using GIF and JPEG despite the fact that they've been known for years to be shite.

Also, WASM supports a binary form, so you can skip parsing.

It would mean that image decoding can’t start until the page JS is initialised, which is a pretty crappy trade off compared to today.

The point of using a format over another, given the availability, is to gain performance.

You're suggesting to load a whole new parser, parse images at a higher level than browsers can (i.e. slower and with less concurrency), and make `<img>` a whole lot harder to manage.

That's not worth it at all unless the savings are substantial (say BMP vs JPEG)

The web we got today are an astounding feat of not caring about performance in the most aggressive way imaginable.

Performance is not a factor in this.

> The web we got today are an astounding feat of not caring about performance in the most aggressive way imaginable.

> Performance is not a factor in this.

If you care about SEO, it does. Check Google's pagespeed test.

If that is true then the web does not care about SEO. And quite obviously it isn't weighted particularly hard at all (probably because there wouldn't be anything left).

edit: Case in point, except for google.com (which itself has a terrible score of 62) I can't even find out a single google service that isn't in the reds on the pagespeed index.

Even when it decided to load the mobile version.

> Performance is not a factor in this.

What a completely ridiculous statement. Why else would anyone choose BGP if not for performance (i.e. download speed)? The choice to use BGP over a plain old JPG would be made specifically by people who do care.

The context here is rendering performance.

>and patent free.

Royalty Free.

Purely from a Web Image usage perspective, Jpeg XL is very competitive in terms of features and performance. Also Royalty free.

Hasn't JPEG XL and AVIF largely surpassed BPG in popularity? AVIF is already in browsers.

https://jpeg.org/jpegxl/

https://gitlab.com/wg1/jpeg-xl

https://aomediacodec.github.io/av1-avif/

https://github.com/AOMediaCodec/libavif

https://caniuse.com/avif

BPG has never been very popular, though the “official” variant (heif which is the same idea by the MPEG) is supported by windows, macOS, iOS and Android.

Yes. BPG (and H.265 it uses) may be technically excellent, but H.265 is a commercial patent-encumbered codec. Patents cover any implementation, even free and open-source ones, so in countries that allow software patents BPG is illegal to use without a patent license for H.265.

"Hope you're not worth suing" may be an OK strategy for many people using BPG on their websites, but it's a blocker for wide adoption of the format. The licensing situation is so bad, that the industry went all behind the AV1 alternative, and the AVIF format rides that wave.

https://aomedia.org/

This is by the same person that wrote ffmpeg.

Discussion 2 years ago: https://news.ycombinator.com/item?id=17587684

Discussion 6 years ago: https://news.ycombinator.com/item?id=8704629

and qemu

and tcc (the tiny C compiler).

Those who can be called genius are few in the world, but Facrice Bellard is certainly among them.

I wish the Computer History Museum would do an "Oral History" video with him. Such a fascinating individual with monumental contributions to tech.

And my axe.

Also https://news.ycombinator.com/item?id=10604465 from 2015. Thanks!

and http://jslinux.org/

> 14bit (...) It gives a higher dynamic range (which is important for cameras and new displays) and a slightly better compression ratio (because there are less rounding errors in the decoder).

That's clever and a bit counterintuitive.

Yeah, I remember reading years ago a paper explaining the same for H.264 and why 10-bit encoding matters (even for 8-bit content). It was fascinating reason, btw.

AFAIK, there is also less banding issues even in 8-bit content, and the filesize is generally smaller too. The biggest reason against its usage is more computational power, though.

Was there really a paper on the topic ? I thought it was x264 mailing list from Jason Garrett-Glaser / Dark Shikari.

Practically superseded by HEIF, which is same idea but in complicated ISO container: https://en.wikipedia.org/wiki/High_Efficiency_Image_File_For...

to wait until BPG becomes an open format would take a lifetime. I've been waiting for years now. Fortunately webp is now widely adopted so no need for waiting for bpg anymore

once upon a time i designed an automated framework to attempt to brute force optimal x264 encoder settings for a small library of sample clips. i used lossless BPG for storing select keyframes from various test runs before analysis. i soon learned that having software try to decide the quality of one image versus another is not a terribly easy problem to solve. thanks for the trip down memory lane =)

Unfortunately its patented.

According to Wikipedia:

> According to Bellard's site BPG may be covered by some of the patents on HEVC, but any device licensed to support HEVC will also be covered for BPG. Patent issues may prevent JPEG replacement by BPG despite BPG's better technical performance.

So it sounds like you’re ultimately correct.

The problem is the JS decode mean images in practice load slower than competing formats.

So it's like WebP but based on HEVC instead of VP8.

Unfortunately HEVC is a patent minefield and unless you have deep pockets to stomach patent trolls, implementing anything with HEVC is a risk.

Of these new image format I am a fan of FLIF

http://flif.info/index.html

FLIF has been integrated has the lossless mode of JPEG XL, the next JPEG standard.

FLIF was superseded by FUIF and then by JPEG XL, but it hasn't been fully integrated. JPEG XL has parallel decode and uses MAANS entropy encoding instead of MANIAC, which results is slightly less compression for some situations (but faster decode). That said, I certainly agree that JPEG XL is already very impressive and hopefully the next majorly supported image format.

but how is it pronounced?

The G is pronounced just like it is in gif.

Don't even start.

Neural compression is the future, not this.

Reply on news.ycombinator.com