HD @ low bitrates for H.264: mythes and reality

[dr.CMOK]

Hi, everyone!

First of all, could not find the subject here, so if it was already discussed, then sorry.

Now, back to subject.

To start with, a small preface. Recently (more or less :) LS announced its 200 series which, as they claim, can do 720p30 at just 768k. I've also seen a discussion at WR forum about it, but, unfortunately, was unable to register there to ask this nasty question.

So, being an engineer, I hardly can imagine that one can do this (720p30@768k) with H.264 (i.e., MPEG4 part 10) without sacrificing quality, e.g. Y-PSNR which is roughly can be treated as quality estimation for H.264 types compression.

Again, it's well known, that you can achieve rather good (yep, good, not excellent) quality when compressing DVD video at that bitrate (H.264 compression), so, if one compares resolutions for DVD image size and 720p, it's easy to get rough estimation of bitrate needed for it with H.264: it's about 5M. Of course, in VC we don't have exploding robots or spacecrafts, still, claiming that 768k is enough for 720p30 in some general case does not look very reasonable, does it?

Just to mention a little bit more, of course, one can claim even lower bandwith (e.g., under condition that the room will be completely dark and there are no sudden spacecrafts explosions) :). Nevertheless, when talking about VC, we have to remain within standards boundaries (H.264 in this discussion). Even more to notice: today's VC is not just a group of big bosses sitting around a table, but also a telemedicine, disaster reporting and recoverng, etc...

[Sean Lessman]

I will link this post over to Wainhouse, you have some valid points.

Sean

[dr.CMOK]

I will link this post over to Wainhouse, you have some valid points.

Sean

Thank you, Sean, it'll be interesting to see reaction out there :)

[Sean Lessman]

Thank you, Sean, it'll be interesting to see reaction out there :)

Posted.

http://www.wainhouse.com/forums/showthread.php?p=15791#post15791

[dr.CMOK]

Well, as I've been guessing, the post deeply touched LS-guys at WR :)

Just a small step aside for not to be blamed for being pro-TAA, or pro-PLCM, or pro-whatever-else: I'm not working for any of those companies, I do aware of most popular VC equipment in Russia, I do have experience in working with some of it. Throughout my projects development process I do not care who is the vendor (unless customer wants it to be some exact vendor) as I've used to build the best solution the customer can afford within his budget, without looking at fancy "we're the greatest" product sheets.

Now, to give some feedback to the replies at WR.

1. schwenn and notice about bandwithes in Germany.

Sitting in the midst of a Gigabit research net imbedded in the open internet, this discussion about minimal bandwidth for HD VC is of course since several years irrelevant for us.
AND even more - for 399 € / month everybody can get a 6 Mbps (flatrate from QSC in Germany) access to the open Internet. Why do people stick to providers like Telecom etc. which may be necessary for worldwide secure data networks? VC over IP with H.235 encryption is save per se, so use the open Internet for VC and stop this anachronistic bandwidth debate.
Best regards
Ulli Schwenn

What can I say - they're lucky to have such a cheap Internet connections, I believe also with the same security and QoS level as VPN/MPLS/METRO/ISDN or others way more expensive ways. In lots of other countries for a customer it's often a big issue to get a resonably priced quality connections, so bandwith conciderations are often vital.

2. smcgee from US

I wasn't aware that H.264 had a hard PSNR (peak signal to noise ratio for the non-engineers here) standard. The only thing I've seen mentioned is a generally acceptable range of 20-40db which is huge.

Well, it's not a standard and I've never claimed that, I just have said that Y-PSNR (not just a some PSNR, which is not really correct) can be treated as rough approximation of quality (well, at this point I've asked few of mine natively english speaking friends if they have understood what I've wanted to say, they have). But what concerns range of 20-40dB - it's far from being truth. Acceptable is range 30-39dB (starting 39 and higher you'll get so called "broadcasting quality"). What this means in true-life world is, e.g., 34dB is approximately what you see on your mobile when watchin 3gp videos (and normaly you don't scale them to 50-inch plasma display).

The Room 200 product can offer an equivalent level of HD video quality at 768kbs due to technical improvements in the product.

What is the level of that equivalence? And does "equivalent" implies, that it's not really an HD?

3. To H323, again from US (I beleive, he or she was asking Sean)

What would your answer to Dr. Cmok be on how tandberg gets to HD at 1.5 mbps when his formula calls for 5 mbps?


Actually, I dont' remember that TANDBERG claims some fixed framerates for HD size (720p) picture for given bandwith in their datasheets.

So my next question to those who disagree with me, how can you keep quality when decreasing bitrate? Normally, all the tools you've got to do this is changing codec complexity, adding additional passes for encoding, changing GOP structure... that's it in first approach.

[Sean Lessman]

Seems you have some constructive things to say, you should try to register to Wainhouse again so you can contribute. Good luck!

Sean

[dr.CMOK]

Seems you have some constructive things to say, you should try to register to Wainhouse again so you can contribute. Good luck!

Sean

No, thanks :) Since first attempt nothing has changed in me, so normaly they shouldn't have any pro-s for doing it. And, actually, I like this forum more. At least, it seems that more people not just writing, but also reading here :)

[smcgee]

So my next question to those who disagree with me, how can you keep quality when decreasing bitrate? Normally, all the tools you've got to do this is changing codec complexity, adding additional passes for encoding, changing GOP structure... that's it in first approach.

You've answered your own question.

-Smcgee

[dr.CMOK]

You've answered your own question.

-Smcgee

Well, I've just provided few pro-s (at least, I beleive, those are pro-s :) ) to my point of view on the truth (which is, as we know, is out there).