New H.265 twice as efficient as H.264!

[singhcr]

With all of the talk of expanded color space and chroma subsampling, I was a bit lost so I did a bit of research.

The color space bit makes sense as you are expanding the range of colors that the display can show. Having 10 bit color vs 8 bit means you can have a more nuanced color display. However, I was really lost when it came to 4:4:4 vs 4:2:0, etc.

So, here are some good primers on the subject.

What is YUV color?
What is 4:4:4 (chroma subsampling)?

[Penton-Man]

Originally Posted by singhcr

What is 4:4:4 (chroma subsampling)?

On that note, in the battle over providing more detailed imagery, I’m looking forward to when the previously mentioned Y"u'v' 4:2:0 signal paper (last paragraph here - https://forum.blu-ray.com/showthread...er#post9596655 is presented and published

and seeing if it at least has a shot at an Honorable Mention in 2014, like these past ones from last year’s IBC….
http://www.theiet.org/communities/mu...a/ibc/2013.cfm

[Penton-Man]

Touching a tiny bit on current state HEVC implementation rather than dwelling only on codec development and incremental steps towards standardization of advanced features, like with the previous posts on the last few pages...

Does anyone know, or can anyone even fathom a guess (given the computing hardware resources available at the studio level or major content provider) as to the present state of HEVC encoding for movies in regards to how long does it actually take to encode a 4K movie (let’s say into 8-bit 4:2:0 4K delivery product) from a high quality master file with HEVC as compared to the time it now takes using optimized H.264 encoders?

No significant difference?
5% longer, 10%, 20%, 2x, 3x, 5x as long?
Or more?

Because as the old adage says…”Time is Money”.

[Penton-Man]

Correct answer is….

’Or more’...but… this is good considering the early lifecycle of HEVC.
Next question….when do you think content providers can get it down to near that of the same time for the x264 encoder?

[Penton-Man]

Originally Posted by Penton-Man

Next question….when do you think content providers can get it down to near that of the same time for the x264 encoder?

Nobody?
Vargo….how bouts an educated guess.

[vargo]

Originally Posted by Penton-Man

Nobody?
Vargo….how bouts an educated guess.

Well you can always do it faster if you throw machines at it. Or dedicated silicon. If we're talking a direct comparison - HEVC software encoding vs H.264 software encoding on the same hardware?

The main problem is that software encoders like x264 are highly configurable.. the slowest x264 preset is dozens of times slower than the fastest. The speed determines the compression efficiency.

Therefore you can already encode x265 faster than x264, if you choose the fastest and slowest presets respectively. Of course it's a rigged test - at the same bitrate, x265 will look much worse, defeating it's purpose.

Unfortunately we're probably going to see vendors do exactly that - use low quality settings to make BS speed claims about their encoder. We're also going to see content providers deliver sub-par HEVC simply because they're not willing to use effective settings (this was often the case with H.264 until quite recently).

HEVC encoding lends itself to parallelisation and that's the direction that computing architectures have been going and will continue to go. So hardware will effectively 'come to' HEVC. I would say in 4 years the machines of the day will probably do what you are asking - ie software encode HEVC at the same speed as the best H.264 encoders while delivering close to the promised efficiency gain.

[Penton-Man]

Originally Posted by vargo

I would say in 4 years the machines of the day will probably do what you are asking - ie software encode HEVC at the same speed as the best H.264 encoders while delivering close to the promised efficiency gain.

Nice explanation ….plus, you already answered the next part of the question I was poised to pose…i.e. “the promised efficiency gain” part of the puzzle.

Anyway, to tease apart the time intervals, I think it will take about 2 years to get the time to encode (equal quality) close to that of x264, at which milestone you will get a 20-30% efficiency vs. x264. Higher encoding efficiency (40-50%) will take 2 more years in order to come to fruition. So, ultimately, I think we reach the same conclusion.

Anyway, so that the bidness thinking folks don’t feel left out of the discussion by only concentrating on the technical aspects of HEVC, here’s a question for dem business types….

Remember this past statement by MPEG LA….. http://www.mpegla.com/main/PID/HEVC/default.aspx

Question: Who are some of the prominent IP owners (companies) not on the list of supporters and potentially what is the significance of their absence?

[vargo]

Originally Posted by Penton-Man

Nice explanation ….plus, you already answered the next part of the question I was ready to pose…i.e. “the promised efficiency gain” part of the puzzle.

Anyway, to tease apart the time intervals, I think it will take about 2 years to get the time to encode (equal quality) close to that of x264, at which milestone you will get a 20-30% efficiency vs. x264. Higher encoding efficiency (40-50%) will take 2 more years in order to come to fruition. So, ultimately, I think we reach the same conclusion.

Anyway, so that the bidness thinking folks don’t feel left out of the discussion by only concentrating on the technical aspects of HEVC, here’s a question for dem business types….

Remember this past statement by MPEG LA….. http://www.mpegla.com/main/PID/HEVC/default.aspx

Question: Who are some of the prominent IP owners (companies) not on the list of supporters and potentially what is the significance of their absence?

You're probably not far off.

When H.264 was introduced, the encoding complexity was a big deal, amatuers and professionals alike complained about how much longer it took.

In the years following, multi-core processors became the norm and H.264 encoders were highly optimised to work well with multiple threads. Nowadays on a modern machine you can actually encode x264 faster than Xvid and still yield better compression. It makes better use of the available cores.

But on a very powerful multi-core system you can still struggle to get 100% CPU utilisation with x264 and I expect that's one area where HEVC encoder developers will be keen to improve.

[Penton-Man]

How about the F/U question? Can you name any of the companies, or would you prefer to hand that one off to the bidness crew?

[Penton-Man]

Originally Posted by vargo

You're probably not far off.

I don't like giving predictions but, it's just something people should be aware of.

[Penton-Man]

Originally Posted by Penton-Man

...so that the bidness thinking folks don’t feel left out of the discussion by only concentrating on the technical aspects of HEVC, here’s a question for dem business types….

Remember this past statement by MPEG LA….. http://www.mpegla.com/main/PID/HEVC/default.aspx

Question: Who are some of the prominent IP owners (companies) not on the list of supporters and potentially what is the significance of their absence?

Answer:

Several prominent IP owners the likes of Dolby, Panasonic, Samsung and Microsoft are not listed in that group. What that means is that if you are an encoder/decoder user, not only will you have to write a check out to MPEG LA patent group, but potentially you may also have to write out an additional check to patent rights holders (some as noted above) which are not listed in the group.

[Wendell R. Breland]

Originally Posted by Penton-Man

Question: Who are some of the prominent IP owners (companies) not on the list of supporters and potentially what is the significance of their absence?

I pointed out sometime back that Sony and Panasonic was MIA in regards to HEVC MPEGLA.

At one time it appeared MPEGLA would handle Blu-ray patents but that dissipated and One-Blue became the clearing house. It appears there is an underlying story of why Sony and Panasonic are not part of HEVC MPEGLA.

[Penton-Man]

Well, one good thing is that unlike for encoder/decoder users (which will have to pay), the members of the MPEG LA group have agreed to not levy a toll specifically in regards to HEVC content, as was the case with H.264.

Perspective: Despite us critiquing some of the challenges and potential pitfalls in the implementation of HEVC at the current time, lurking readers should not get the impression that all is doom and gloom in the world of HEVC. On the contrary, at least one BIG player (content provider) is jumping in feet first with investment as to its usability because they appreciate its great potential. Not to mention the fact that most all new development is now geared toward HEVC.

[Penton-Man]

Scientific ‘shootout’ performed by the EPFL (https://forum.blu-ray.com/showthread...fl#post8587823) between three contenders….HEVC (HM 15.0), VP9 and AVC

encoding ‘4K’ material (well, to be exact, see slide #5 in the link at the end of this post).
http://soundbible.com/56-Boxing-Bell-Start-Round.html Annnnnnnnd the winner is....

HEVC (skip to slide #18)
For dem slides, see - http://www.slideshare.net/touradj_eb...014-hev-cvsvp9

[Penton-Man]

As a requirement of the above ^ scientific codec shootout, all observers were screened for visual acuity (with Snellen charts e.g. see last page of this pdf for chart samples offered with a commonly used projector for this purpose - http://marco.com/brochures/cp670.pdf)

as well as for observer color perception deficiency with Ishiara (http://www.colour-blindness.com/colo...r-test-plates/).

[Penton-Man]

Originally Posted by Richard Paul

DCI P3 getting chosen for UHDTV would be a victory for the studios, since they wouldn't have to upgrade their equipment, but in my opinion it would be a defeat for consumers since the color space chosen for UHDTV will be used for decades.

The LCD display that Nanosys showed at SID could show 91% of the Rec. 2020 color space and if it had better color filters it could have shown 97%.

Richard, in case you can’t make it to the annual SMPTE Tech Conference and Exhibition in Hollywood this coming October, there will be a presentation including some of the salient points regarding the chromaticities of non-monochromatic light sources and the gamut coverage of real object colors by assuming QD-LED backlit displays with color filters for realizing BT.2020 from this recently published paper -

http://www.opticsinfobase.org/view_a...bile%3Dno&org=

[vargo]

x265 psycho-visual features are now completed and really help a lot with grain retention and fixing the softness problem that affected x265 in months gone by.

In this example we can see less artifacting and better grain retention at 75% bitrate. Quite impressive.

http://screenshotcomparison.com/comparison/90086

Of course the x265 encode took me about 5 times longer with these particular settings, as to be expected.

[Penton-Man]

What would that be by now, Ver. 1.3? Well done , shall we assume the improvement in appearance is at least visually sustainable or even superior when you viewed that same content in motion?

For a heads-up, John Pallett from Telestream (who co-founded the x265 project) and spoke at the last NAB 2014 providing practical data comparing HEVC with MPEG-2 as well as H.264 and suggested recommended HEVC encoding settings for encoding of SD and HD content, will be speaking at the upcoming SMPTE 2014 Annual Technical Conference & Exhibition in October and present new data (since NAB 2014) regarding recommended bit rates and settings for the encoding of 4K (UHD) content.

[vargo]

Originally Posted by Penton-Man

What would that be by now, Ver. 1.3? Well done , shall we assume the improvement in appearance is at least visually sustainable or even superior when you viewed that same content in motion?

In motion nothing changed - x265 looked significantly better at 75% bitrate.

The key point is that there is finally good grain and detail retention. Due to the strong blurring effect there was really not much point in performing in-depth tests or comparisons prior to now.

[vargo]

The iPhone 6 specs mention H.265 facetime. It would suggest that their newest SOC has HEVC hardware encode/decode. It's not mentioned elsewhere so we'll have to wait and see.

If this is true, I think there is a good chance their next set-top box will feature HEVC decode also.