Alliance for Open Media codecs

[HerpaDerp]

Originally Posted by LigH

Imagine, I once found a website which specialized in offering images in WebP format and was not just a format promoting site.

One.

User acceptance is hard to predict. If it doesn't offer more important advantages than just bandwidth saving (in times of VDSL and 4G), there may not be much incentive to use it. Just remember the overhead of external JavaScript modules required to keep your own source small.

I've been improving a website for a customer off/on for the past year. One of the goals was to create product pages that showed high resolution, high quality images with alpha transparency overlayed on top of a fixed background image, so that it makes the product "pop" and look impressive when the user scrolls down the page.

I quickly ran into a roadblock where, JPEG wouldn't work because it doesn't support alpha transparency. GIF will get you close, but it's either fully opaque, or fully transparent. PNG in my instance resulted anywhere from 800KB to 1.5MB per image, and with some pages containing 20+ images, you can see how this would be a problem for mobile users.

So, my solution was to provide WebP to Chrome users on both Desktop and Android, and provide a fallback of Quantized (Lossy) PNG for Firefox, Edge, Internet Explorer, and basically any browser that doesn't support WebP.

This resulted in a massive bandwidth reduction for the user, which caused the page to load 10-15 seconds faster on 4G.

Now, you mentioned you once found a website that specialized in offering images in WebP, and I can name a big one right off the top of my head: Netflix. They in fact, provide WebP to Chrome users.

Facebook at one point dabbled with WebP, but due to the nature of users wanting to download pictures, I think they ended up upsetting people more than making them happy, since WebP is a bad format to serve to people wanting to make local backups of their files.

Now, WebP is really not the best format, but it solved a big problem for me on Chrome, especially since it captures more than 60% of the total browser marketshare. I'm not making that number up.

A big problem with WebP is that it suffers from generation loss, and for certain situations, it can produce a significantly worse looking image than JPEG, which is why it's paramount that you don't choose a quality lower than 80%.

I looked into FLIF image format, which as of right now, is supported by absolutely nobody. It's some new fangled format that shunts the most important bits to the the beginning of the image so that if you downloaded a mere 10KB of a 2MB image, you could still view a low resolution version of it, which quite frankly is awesome.

The see several immediate problems with FLIF.
1. Files take a horrendous amount of time to create if they're greater than 2048x2048px in dimensions because it's doing all sorts of color sorting and algorithms to achieve the best filesize. Time to create the file increases exponentially with dimension. We're talking 5+ minutes to create a 4096x4096 FLIF image. Also, I have been unsuccessful in creating 16bit FLIF files, but that could just be a limitation of XNViewMP. Not sure.

2. It doesn't actually solve filesize problem for mobile users. Hypothetically speaking, if you put 30 FLIF files on your web page, each of which amounts to 2MB in file size, then you're forcing the user to download 60MB of pictures. The only benefit they get is the ability to view the image before they've finished downloading, but this just simply inflates the filesize of the page enormously, and causes other files to queue during the page load time. At the end of the day, serving your users smaller images ( less than 1MB in filesize) is the absolute best way to achieve fast page load time.

3. Animated FLIF is fantastic if you have less than 30 second recordings, but if you go beyond that, it quickly becomes impractical because the user would still have to download a mega amount of data just to be able to begin displaying the image. So, at the end of the day, you're still far better off using a video codec for anything exceeding 30 seconds.... and we now have an absolute wealth of services that do exactly that (imgur, gfycat, etc). Animated FLIF is a neat proof of concept, but I see almost no practical use for it other than being able embed fast previewing short animated images into a product page... which tbh, I would probably be the one in 100,000 website developers taking advantage of that feature. lol... Right now, my current method of doing animated video is to overlay a VP9 WebM for Firefox and Chrome users, and then fallback to animated GIF for Edge / IE users. What a pain the neck, but hey it works.

4. A 30KB partial download of a FLIF looks SIGNIFICANTLY worse than WebP, Jpeg, BPG, or any new fangled image format that is saved at the exact same filesize. FLIF is fantastic looking once the full data has been received, but looks quite terrible for partial downloads if you're comparing them to already existing image formats saved at the exact same filesize.

5. If web browsers ever become intelligent enough to download "only" the bits needed for optimal viewing experience, what we will end up with, is a bunch of partial downloads sitting in our browser cache, which means if user resizes their web browser, or zooms the image, then the browser will have to request the additional data from the webserver, which means, you can end up having multiple requests for the same image... and web servers operators HATE multiple requests. They would rather you get the full file the first time, than to bog down their servers with multiple requests for the same file.

I could go on and on and on.

But what we need, is not FLIF. We need an alternative image format, that reduces file size, has features, does not suffer from generation loss, is fast to create files, and in every way shape or form, replaces JPEG and PNG. FLIF does not replace JPEG and PNG. It's great, but it doesn't resolve the 5 issues I mentioned above.

And unfortunately, AV1 probably has TERRIBLE generation loss because using similar tech as WebP, and I can almost guarantee, it will suffer from the same problems. Nevertheless, if an AV1 image format became available to use in Chrome, I would switch to it simply because I like better things.

Anyway, just thought I'd like to share. lol

[Jamaika]

Originally Posted by HerpaDerp

I quickly ran into a roadblock where, JPEG wouldn't work because it doesn't support alpha transparency.

We are talking about the JPEG standard. What about the other native JPEG development?
JPEG2000, JPEGLS or JPEGXT. Amateur production or codes will have any application.

Originally Posted by HerpaDerp

Now, WebP is really not the best format, but it solved a big problem for me on Chrome, especially since it captures more than 60% of the total browser marketshare. I'm not making that number up.
A big problem with WebP is that it suffers from generation loss, and for certain situations, it can produce a significantly worse looking image than JPEG, which is why it's paramount that you don't choose a quality lower than 80%.

Google develops a WebP but without the VP9 codec. I don't know what creators have vision Lepton. Codec bases on WebP and JPEG. I think that you made a mistake codecs. WebP isn't 16bit, and 16bit is FIiF. Unfortunately, it works even more slowly.

[HerpaDerp]

Originally Posted by Jamaika

We are talking about the JPEG standard. What about the other native JPEG development?
JPEG2000, JPEGLS or JPEGXT. Amateur production or codes will have any application.

https://caniuse.com/#search=jpeg

Google develops a WebP but without the VP9 codec. I don't know what creators have vision Lepton. Codec bases on WebP and JPEG. I think that you made a mistake codecs. WebP isn't 16bit, and 16bit is FIiF. Unfortunately, it works even more slowly.

[benwaggoner]

The only real viable replacement for GIF/PNG/JPEG I see now is HEIF, which is now the default picture format for iOS. The implementation today is HEVC, but extending it to AV1 appears reasonably straightforward.

I have no idea how HEVC and AV1 would compare for still image coding, but both are WAY better than JPEG or PNG for any kind of image type. And unbelievably better for mixed continuous/discreet tone images like graphics or comic books.

[hajj_3]

Originally Posted by benwaggoner

The only real viable replacement for GIF/PNG/JPEG I see now is HEIF, which is now the default picture format for iOS. The implementation today is HEVC, but extending it to AV1 appears reasonably straightforward.

I have no idea how HEVC and AV1 would compare for still image coding, but both are WAY better than JPEG or PNG for any kind of image type. And unbelievably better for mixed continuous/discreet tone images like graphics or comic books.

isn't heif patented though? I can't see mozilla adopting a patented format nor many phone manufacturers. AV1 is 15% better than hevc at still images according to recent statements. Google's PIK is another contender.

[mzso]

Originally Posted by hajj_3

isn't heif patented though? I can't see mozilla adopting a patented format nor many phone manufacturers. AV1 is 15% better than hevc at still images according to recent statements. Google's PIK is another contender.

I don't think Pik can exist if they go for an AOM image format. If they decide to diverge from AV1 compatibility it's improvements will probably be merged, otherwise it'll likely get abandoned.

[IgorC]

Originally Posted by benwaggoner

The only real viable replacement for GIF/PNG/JPEG I see now is HEIF, which is now the default picture format for iOS. The implementation today is HEVC, but extending it to AV1 appears reasonably straightforward.

Can I ask why do You think that HEIF is the only viable alternative?

Not sure about HEIF but HEVC has zero chances to replace jpeg.
Enough to say that Google and Mozilla simply won't implement it. It's a big no-go. There is no even a minor doubt of that.

JPEG XL or intra-AV1 have all chances. Or some sort of collaboration between these two.

[mzso]

Originally Posted by IgorC

Can I ask why do You think that HEIF is the only viable alternative?

Not sure about HEIF but HEVC has zero chances to replace jpeg.
Enough to say that Google and Mozilla simply won't implement it. It's a big no-go. There is no even a minor doubt of that.

JPEG XL or intra-AV1 have all chances. Or some sort of collaboration between these two.

Well, HEIF is HEVC based, so I doubt there's any difference in the corporation's attitude toward it compared to HEVC.

[dapperdan]

Originally Posted by IgorC

JPEG XL

"To encourage widespread adoption, an important goal for this standard is to support a royalty free baseline."

Sounds like a non-starter with wishy-washy language like that. Pretty sure that's what was said about H.264 as well, but they had no real process in place to make it possible. HEVC even lacked a process to enable one-stop shopping for royalties. They all seem to be stuck in some trap of their own making.

[IgorC]

It's just a beginning.

"The JPEG Committee intends to publish a final Call for Proposals (CfP) following its 78th meeting (January 2018)"

There was already some collaboration(?) between Daala developers and JPEG
https://people.xiph.org/~jm/daala/revisiting/

The improvement in quality compared to the previous status update is quite obvious. Daala is now much better than both WebP and JPEG (libjpeg-turbo is the most commonly used JPEG encoder). As for Daala vs BPG/HEVC, the artifacts are obviously different and hard to judge from just four images. Opinions are likely to vary based on the viewer and the input image. At this point, what we'd really need is a full subjective test. Fortunately, this is exactly what is going to take place shortly, as Daala has been submitted as a candidate for the Image Compression Grand Challenge at ICIP 2016. The results should be available in September. In the mean time, you can read the paper we will be presenting.

I guess that now Daala is part of AV1 project there will be some feedback between AV1 and JPEG-XL.

[mzso]

Originally Posted by hajj_3

new Libvpx released with improvements to VP9: https://www.phoronix.com/forums/foru...mizations-more

This is the AOM/AV1 thread. There's vp9 thread also.

That said, I hope there's some multi-process optimization. CPU utilization is abysmal when encoding VP9.

[GTPVHD]

http://blog.chiariglione.org/a-crisi...nd-a-solution/

At long last everybody realises that the old MPEG business model is broke, all the investments (collectively hundreds of millions USD) made by the industry for the new video codec will go up in smoke and AOM’s royalty free model will spread to other business segments as well.

Even the MPEG founder sees the beginning of the end of MPEG's terrible proprietary codecs situation & business model.

[birdie]

Originally Posted by GTPVHD

http://blog.chiariglione.org/a-crisi...nd-a-solution/

Even the MPEG founder sees the beginning of the end of MPEG's terrible proprietary codecs situation & business model.

Thanks for the link.

HEVC is 60% more efficient than AVC? I wonder what he's smoking. In my experience it's more like from 0 to 30% (for very specific use cases like anime).

[LigH]

In UHD resolutions it's more probable for HEVC to be ahead of AVC.

[Mystery Keeper]

I've only ever encoded one video with HEVC. A musical video with rather high motion. At the same bitrate HEVC did visibly much better than AVC. Also, took forever to encode.

[TD-Linux]

MSU's codec comparison is out: http://www.streamingmedia.com/Articl...ticleID=122945

It uses a version of libaom from July 2017.

[Selur]

btw. is there any news regarding when to expect a bit stream freeze? Last I read was January 2018, which now is gone,..

[mzso]

Originally Posted by Selur

btw. is there any news regarding when to expect a bit stream freeze? Last I read was January 2018, which now is gone,..

It's not midnight yet everywhere. :P

[Rumbah]

Originally Posted by Selur

btw. is there any news regarding when to expect a bit stream freeze? Last I read was January 2018, which now is gone,..

FOSDEM is this weekend so I guess that's when we get some news.

[iwod]

3000x slower.

Also we need to stop focusing on PSNR and SSIM. These two should be used to add prospective.

Last time i heard Netflix were going to update VMAF. What happen to that?