New video-compression (HEVC) twice as efficient as current standards.

[ckmox]

Originally Posted by JEEB

Anyways, the "next generation formats" I see right now in development are:

• HEVC (developed in the usual way that we know from the earlier standards)
• vp8-dev/vp8-next (yes, this is why Google really doesn't care about VP8 that much, if at all; developed at Google behind closed doors IIRC)
• Daala (openly developed by Xiph so that anyone can join the effort)

just a little off-topic concern... can anyone make a dedicated thread for Daala too?

we already have dedicated threads for VP8-Next (NGOV) and H265 (HEVC)

[Parabola]

Originally Posted by Keiyakusha

I wouldn't be so sure because HEVC's main advantage is higher compression, but this comes from increased complexity and even decoding will require more CPU power. Since current hardware hardly use all neat h264 features due to performance reasons, It doesn't looks like something even more complex is what they really interested in. (Edit: unless Apple wants the video to drain batteries in their i-stuff faster). Most likely it will face the fate of mpeg4-asp until blurays will be switched by some new purple-rays or whatever-rays and yet another standard will be developed.

HEVC will be bigger than MPEG-4 pt. 2 ASP. Decoder complexity is not as bad as you might expect and HEVC is already heading straight for silicon. And, you'll save battery in running less bits over the new DSP-heavy radio protocols.

What could kill it is uncertainty over licensing, or expensive or badly thought out license terms - are you listening, MPEG-LA? Or uncoordinated, multiple licensors. IETF is making interesting noises, but the timing of their initiative is interesting at a point where HEVC patent strategists have yet to reveal their position.

[JEEB]

OK, bumping the general HEVC thread as it seems like the libavcodec decoder project has gotten into somewhat more usable state

After months of getting my samples fail to decode because of the parser ending up in an eternal loop somewhere, it can finally decode my HM 10.0 encoded sample, and going through the per-frame md5 hashes given out by the decoder against the encoder's decoded frames' md5 hashes, it seems to be generally if not fully correct (I cannot be sure of full correctness as I haven't removed the other data that framemd5 outputs in addition to the actual hashes, so I haven't computer-checked them all to be completely exact yet).

That said, the sample was intra-only, as I never got to making other ones until it decoded fully and correctly. P and B frames have seen some development in the decoder, but I would still guess that things aren't exactly fully ready on that front yet (although I could be incorrect). Happy testing, and do remember that you will want to use HM 10, or newer, to encode (or any encoder of your own that you might have access to that confines to the current bit stream standard)!

[JEEB]

Originally Posted by JCT-VC Mailing List

To all the experts of the JCT-VC,

I am very proud to announce that the editors have finalized the HEVC specification text and submitted it to ITU-T and ISO/IEC for final approval (Last Call in ITU-T and FDIS ballot in ISO/IEC).

The completion of HEVC is surely the most significant event in video coding in a decade. It has been an extremely active, constructive, and collaborative project, with more than 750 participants and more than 5500 contribution documents, and I believe the result is extraordinary.

On behalf of the editors and the management team of the project, I would like to thank all the contributors.

Best regards,

The editing team (Benjamin Bross, Woo-Jin Han, Jens-Rainer Ohm, Gary Sullivan, Ye-Kui Wang and Thomas Wiegand)

And thus we have come to this, seems like Draft 10 Version 34 is the version that was sent out to the FDIS ballot. If I recall correctly, an FDIS ballot should take two months, and if it passes (most probably it will), HEVC will finally officially be a real standard

Many people have been working on implementing this format for quite some time now, but I think at this point we can congratulate everyone related to the HEVC project, and thank them for their good work at finishing up the format.

[JEEB]

Ladies and gents, we have our first official spec. While I haven't been looking, it seems like ITU-T ratified the HEVC format as ITU-T H.265 (04/2013) about a week ago.

As the ISO/IEC FDIS ballot takes two months, that will probably take until the latter half of May to finish

[Nil Einne]

Originally Posted by Atak_Snajpera

if opus can be slightly better than aac so i see no reason that people who are behind x264 could not create free alternative to h.265. with proper finacial support from google they may create something alot better than crappy vp8 codec.

I know this was an old post but I keep seeing these comparisons and it seems to me they're poor comparisons. If I'm reading correctly, the Opus bitstream was frozen Jan 2012 and standardisation was completed sometime that year. I don't know about the bitstream but standardisation for HE-AAC was completed in in 2003 or 2004. LC-AAC was even earlier. In other words, we're talking about a 8 year or so gap. It shouldn't really be that surprising that they can finally beat. While HE-AAC codecs have been improving in that time, ultimately they're limited by the spec. And looking at Opus's own info http://www.opus-codec.org/comparison/ it doesn't look like Opus is that revolutionary as some claim. It's better but not by that much. The bigger advantages are probably latency and working range.

In the case of H.265/HEVC vs Daala or VP9 we're talking about each one trying to be the best they can be within certain parameters developed based on current tech level (well VP9 seems a bit behind? and Daala perhaps even more, but that's not necessarily a good thing). So while I wish those projects well, I'm not convinced they're really going to achieve truly equivalent quality. In any case, I suspect the battle will be lost elsewhere namely in the lack of hardware, vendor and content provider support. (As I mentioned, it seems H265/HEVC is going to beat them in terms of being first which in this case considering the legacy and support it has behind it means it'll probably have won before VP9? or Daali really get off the ground.) Of course Daali and VP9? will hopefully end up being better than H264, but that's not a major victory and shouldn't be their goal.

BTW, don't get me wrong. I'd love to see an patent free codec win, it's just that I don't think it's going to happen from all that we've seen so far and all that I'm seeing now. (I would prefer Daali over VP9.)

I should mention I don't know much about Opus, I only learnt about it recently. I'm not dissing it, in fact I'd love for it to become the new defacto standard, the fact that it seems to handle a wide range of bitrates well seems particularly nice. Sadly I don't think this is going to happen. Despite the decent legacy behind Opus, I think the audio codec world is just a bit too dead. It may take over in some applications like realtime audio (including realtime video conferencing, VoIP etc) but in many areas, people aren't going to see a reason to move since the benefit is too small. Enthusiasts may move, those who aren't convinced they need lossless. But in most areas, people are happy with the size/quality tradeoff or just don't care. And there are other problems, e.g. in the multichannel world, we still have DTS and DD5.1 although I presume other codecs could achieve a better quality/size tradeoff. I've never looked in to it, but I've always presumed the reason these formats still dominate is because of the vagaries in streaming other formats to receivers (HDMI and LPCM should negate this but I'm not sure what level of support there is). And of course in the Bluray world, there are a bunch of stuff which likely add nothing but extra space used (which doesn't matter so much for Blurays, but still...). Which is another reason I forgot to mention for the prevalence of the other formats, marketing.

[mandarinka]

Daala might look interesting, but it is still in the conceptual stage (AFAIK they are still designing transform and intra prediction, inter prediction hasn't been on the programme yet), and the development isn't progressing exactly quickly.
In other words, if it will be finished, it is going to come few years from now (by that time, H.265 will likely have formed a reasonably mature environment).

Add to the problem that upon finishing of the specification (and upon review/standardization, should it be submitted to some standard body for that purpose), there will still be the task of writing and tuning a production specification of the encoder. And I think that most of us realise what a gigantic task it is to implement a high-quality video encoder.

[hajj_3]

daala will probably be axed, development is ridiculously slow, it would take years at this rate for a final version to be released. VP9 will probably be released this year, unfortunately the quality is vp9 is very poor at the moment, in many instances beaten by x264. I really hope google doesn't finalise it anytime soon, it needs to be as good or better than h.265 in order to succeed.

[sneaker_ger]

Originally Posted by hajj_3

VP9 will probably be released this year, unfortunately the quality is vp9 is very poor at the moment, in many instances beaten by x264.

Not being beaten by x264 in every instance despite being in an early stage is a darn good performance, I'd say.

[xooyoozoo]

VP9 will probably be released this year, unfortunately the quality is vp9 is very poor at the moment, in many instances beaten by x264. I really hope google doesn't finalise it anytime soon, it needs to be as good or better than h.265 in order to succeed.

Using PSNR (well it's not like they developed for anything else) as a measurement, HEVC barely changed its quality per bit since early 2012. Since mid-late 2011, it might have increased bd-rate by roughly 5-10%.

My main point here is that Google will need more than the remainder of this year to catch up to HEVC. Even with open collaboration from a cornucopia of researchers and corporations, as well as free reign to uses as many patents as needed, progress on HEVC's quality wasn't exactly overwhelming. It seems like the majority of its performance was already determined by foundations laid previously, and I don't expect VP9 to buck that trend by gluing on a bunch of new features in a short amount of time.

In terms of subjective quality, the gap between HEVC and VP9 is still quite distinct. It's obvious enough that, recently, I haven't bothered to try to quantify it with metrics.

[foxyshadis]

Originally Posted by xooyoozoo

Using PSNR (well it's not like they developed for anything else) as a measurement, HEVC barely changed its quality per bit since early 2012. Since mid-late 2011, it might have increased bd-rate by roughly 5-10%.

HEVC tools were actually finalized based on subjective viewing, not PSNR. Most of the work since early 2012 has been nailing down the bitstream and bugs, rather than releasing an incomplete or vague spec, like the VP* series, but room for optimization is massive, once other projects start marrying their H.264 mode decision and optimization algos with the new HEVC spec. VP9 is probably a full generation behind, like VP8 was compared to H.264.

[JEEB]

DivX / MainConcept (now under Rovi) has posted the first RFC for HEVC-in-Matroska.

The RFC is available here, original changeset for mkvtoolnix is here, and their first post on Matroska-devel is here. It is expected that they will soon reply to this thread, and thus re-start the discussion on HEVC-in-Matroska once again .