MPEG2 vs MPEG4  

Initially, Blu-Ray movies will be encoded in MPEG-2 format, and they will slowly move towards MPEG-4. If done correctly, the picture quality of both formats should be the same. The goal of MPEG-4 is to provide the same quality image as MPEG-2 with less bandwidth (primarily for Internet distribution). I personally don't think MPEG-4 is there yet, so I'm glad they are sticking with MPEG-2 a bit longer.

when a movie is transfer to digital it has a certain number of sampling points, these sampling points give the resolution and the frequency. Now these numbers give an extremely large amount of data points. Different algorithms end up throwing a different amount of data away. Over time we learn what is less important and what is more important. As the people that create these codecs learn more stuff they create better codecs and encoders. The issue at hand is that people found out how to make stuff more efficient and came out with better codecs. In theory it means they throw away a lot more of the less important data and keep more of the more important data and so in the end something that will look as good takes up less space. So someone can go with extremely high bitrate MPEG2 and extremely low bitrate MPEG4 and have a higher PQ Mpeg2 then AVC or even VC1. At this point in theory on average at a bit over 1/2 that of MPEG-2 AVC and VC-1 should look as good.

As for what will be used with BD. The answer is that a studio has a choice to use MPEG-2, AVC or VC-1. Sony said they will start off with MPEG-2 because they have plenty of technicians and equipment to do the job right now, while if they went with any of the newer codecs they would not be able to push out the movies as fast. As for the other studios, nothing is known

MPEG4 hd dvd will have more compression artefacts compared to MPEG2 high bit rate B RAY.

it seems blu ray will be the quality HD while hd dvd will be the "cheaper" hd.

Did you actually see the two demoed or is it your opinion based on someone else's opinions? Just curious how you could be so certain...

Hong.

Download any of the apple H264 trailers, crank up your brightness! You'll see a ton of artifacts, esp on the black levels! I know the trailers may be a bad example, but there are still there and it puts doubt in my mind. Sony have said at this time the quality of MPEG 2 encoders is better than any of the VC1 or MP4 encoders.

The thing is, they are a bad example. They are encoded at a rather low bitrate to make downloading easier.

must agree with most here there is no reason why one codec can't look as good as the other assuming they each have enough bandwidth

I agree that that tend to be so. However, that's not generally true, though. That is, one can find some video codecs whose algorithms (e.g., different profiles) have limitations no matter what the available bit budgets are.

That being said, MPEG-2 and H.264 will probably give about the same (that is, percetually indistinguishable from the source) quality given enough bits. However, in real life, you have to live with some constraints...

Hong.

I've been working with Video Codecs from before there was DVD - in fact, produced the first film ever produced and distributed digitally for the PC.

Comparing MPEG-2 to MPEG-4 for quality is missing the point.

MPEG-2 quality it much better than MPEG-4.
The purpose of MPEG-4 is to create a watchable image at extremely low bandiwdth. I have been saying for a while that this is the WRONG codec to use for HDTV. In fact, here is how the codecs were designed.

MPEG-1: Low Processing Requirements
MPEG-2: Quality SD Processing
MPEG-4: Low Bandwidth Requirements
VC-1: HDTV Processing

Of these, only VC-1 was really designed with HDTV in mind. It produces a much more accurate colorspace than any of the others as well as a higher quality image. It may not compress as well as VC-1, but it is better than MPEG-2 as far as compression.

I, for one, certainly hope that VC-1 becomes the de facto standard for HD on these new format discs. It is a significant step up.

I think you're confusing MPEG-4 simple profile (ISO/IEC 14496-2) with MPEG-4 Advanced Profile (ISO/IEC 14496-10, aka MPEG-4 Part 10, aka AVC, aka H.264). MPEG-4 SP is what most on the web are familiar with, in the form of DivX, XviD, etc. MP4 SP was designed to provide acceptable quality at very low bitrates. And I belive it's been shown a few places that MP4 SP doesn't scale well to high resolutions and high bitrates.

Next generation DVD will not use MPEG-4 SP, it will be using H.264/AVC MPEG-4 Part 10. When people say MPEG-4 in regards to next-gen DVD they're talking AVC. AVC by all accounts scales just as well to high resolutions/bitrates as VC-1. Take a look at the monster thread (search for FRExt or JVT and post by and around Tom McMahon).

Haven't kept up too well with MPEG-4 then. What is its relation to MPEG-4 SP? Why the same moniker if the core codec is not the same?

I have played a bit with VC-1 and it is an outstanding piece of work.

I've downloaded sample HD videos from a TV Capture community I belong to. Someone encoded a Conan skit (raw HD .TS stream) to mpeg-4 H.264 (using QT Pro).

It looked amazing, noticably better than the h.263 derivatives. But the strain it takes on your CPU during playback is heavy. I have a P4 oc'ed @ 3.4ghz and going back & forth on the timeline is a PITA. Hardware h.264 acceleration is sounding better all the time.

Again, it's silly to compare D-VHS quality mpeg-2 to made-for-the-web quality mpeg-4. Apple turns the bitrate way low (compared to D-VHS/BR/HD-DVD) to save on bandwidth.

the core is the same, but the different profiles are tweaked and have different algorithms added for different functions

Anthony, your explanation, though valiant :), is not quite accurate on benefits of advanced codecs. That is, a codec like VC-1 doesn't always get better because it throws out stuff you don't see. Sometimes, it flat out does a better job than MPEG-2 without throwing away anything more. Case in point: the "entropy coder". The back-end of all of these codecs is a *lossless* compression engine. Since this part is lossless, no data is thrown away regardless of codec used. Yet, VC-1 uses a more advanced algorithm that can be 10 to 15% more efficient than MPEG-2's entropy coder. So you get something for nothing :).

Other gains also come from using variable block size where even better edge definition is achieved than MPEG-2, since the block can fit the source image better than 8x8 blocks in MPEG-2.

All in all, this stuff gets complicated quickly :). But the net summary is that VC-1 can always match the picture quality of MPEG-2 but definitely not the other way around. By using MPEG-2, BD leaves quality on the table for no reason, especially in BD-25 variant which will be used for now.

Amir
P.S. Thanks for the kind words Peter :).

Didn't know they were "kind words". But I assume from your location, you work for Microsoft.

Many, many years ago my company (which was actually myself and two others at the time) produced the "It's a Wonderful Life" CD-ROM. In doing this, we were determined to use the then new "Video for Windows" system. Problem was that it would not encode video longer than approx 10 minutes (due to a 16-bit limitation on number of frames if I remember). I remember working with Microsoft on new algorithms and encoders to enable us to not only encode the entire movie, but to add features in allowing us to "sync" the script with the movie, insert additional data tracks, and much more.

It is now many, many years later, and while I have a new and different life, I see that Microsoft has grown the single programmer we worked so closely with into an entire group which has produced some amazing code in VC-1.

I suggest everyone check it out for themselves. Do a side-by-side with MPEG-2, and MPEG-4 (used Apple's variant) and see what you think for file size vs. quality vs. decode overhead. I think you will be impressed and surpised by the result.

Sorry Peter, I'm not up on the technical details, I just know that H.264 (MPEG-4 AVC/MPEG-4 Part 10) is not the same animal as the "web" variant of MPEG-4 (Part 2). It's relatively new in terms of widespread availability and the tools are still maturing.

I'd suggest you play with Nero Recode some and be sure to use the AVC profiles. :) If I remember right, Apple's implimentation of H.264, at least in their consumer Quicktime version, is not that great. Definitely not optomized for PC playback.

Used Nero Recode a lot. Tried it out for a variety of uses. It is an excellent codec and had been recommended to me from a variety of sources.

However, looking at the two in a couple of side-by-side comparisons (Nero and VC-1) the differences are not subtle. VC-1 holds a richer color palette - as shown in variance of tones of the primaries and it also does a better job at holding a unified field (ie a field of a single tone).

I can only surmise that Recode does a culling of perceptual colors which throws away some dynamic range for space, at least more than VC-1 does. Also, I am unsure as to the colorspace Recode uses, but it does not appear to be ATSC - as does VC-1.

As far as "edge effects" and "macroblocking" they both appear to be on par. It is in color reproduction that there is a noticable difference.

Amir, like you said

I was just trying to simplify it a lot. don't know if I managed to reach my goal, but I did not want to get into all the details and tricks used in codecs to cut the amount of data used for a given scene.

assuming a fair test, I would agree. But if VC-1 was at 3mbps and mpeg2 at 30mbps something tells me the mpeg2 would look much better for most scenes.

Isn't that the question? what is the difference between two codecs in PQ? the answer is they can look as good or not and will depend on other factors such as bitrate and the scene itself.

10x difference in the bitrates is too extreme a difference, that's not the case here.

every article about HD and BD mentions the "higher capacity" of BD, in GB. but if PQ is more or less equal, time is how capacity should be measured. and the codec differences even that out quite a bit. given that, most articles are misleading on the topic of capacity.

whether PQ is equal is a matter of debate, but odds are PQ on HD will be good enough for most people, which will no doubt bug a lot of people here.

how has this become a HD-DVD/BD thread? the question was simple

the answer is simple (even if not much of an answer) PQ is not dependent on the codec but the bitrate used. Yes some are more efficient but if one looks only at PQ and at nothing else then there is no reason why one must have better PQ, and as an example I picked 3 and 30 because no one would argue that MPEG-2 at 30 would not look better then VC-1 at 3mbps. now if they were both 3 I am guessing VC-1 would look much much better (I have not worked with VC-1, but I know that AVC vs MPEG-2 there would be no comparison even on simple scenes) if they where both 30mbps, my guess is AVC and VC-1 would look a bit better then MPEG-2 but if AVC/VC-1 where at 3mbps and MPEG2 was 30 there is no doubt what would look better. So without knowing all the facts no one can say PQ if mpeg2 is better or worst then AVC or VC-1 ( for anyone that might not know AVC is an other term for MPEG-4 it is also known as h.264)

Well Anthony, we know the approximate bit rate of BD titles. We certainly know they are not 30 Mbit/sec. Because that would be 13.5 Gigabytes/hour so even 2 hours of HD content would push it above BD-25's limit and this is for a silent movie!

As I and others have reported, the data rate Sony is using is in the 15 to 17 Mbit/sec. 2 Hours of HD content at 17 Mbit/sec is about 16 Gigabytes leaving room for a couple of lossless tracks and extras. At this data rate, MPEG-2 is only 30% or so higher than VC-1 so I don't see how you can keep defending the MPEG-2 choice.

Amir

Amir how did Sony just get into this discussion? the question was

now if you and Buzz want to congradulate each other why HD-DVD is better, go ahead. I am not discussing HD-DVD/BD in this thread. It was a clear question asking what the hell all this codec stuff is all about and that is what I answered and where I am limiting my discussion

No the first line was:
"does anyone know how the new blu ray discks will be encoded?"

And the forum we are discussing this in is "Blu-ray software" and not some codec forum. As such, it is very relevant to put the question in context.

You mean as opposed to sweeping the issue under the rug with talks of MPEG-2 data rates people won't see on these discs? :)

No one is discussing other HD DVD/BD topics. We are talking about relevant data for codec choices here. You went off to defend MPEG-2 with data rates that are not in play here. So I had no choice but to bring you back to reality of the situation and why the question was asked.

Amir

well if it is the "Blu-ray Software" and that means questions are automatically about BD then why assume I am comparing to HD-DVD? that would be a question for the general forum. And so assume I am talking about the same info on two different BD disks. If you want to assume that what I said was an HD-DVD bash because the discussion was what looks better and I said you cannot say because the bitrate is important, then I don’t think anything I can type will make you understand that there was no such implication, considering I typed as much several times already.

See post #4 above. :)

I think the difference in arguments is a theoretical configuration vs a practical one. Sony's published info on the BD bitrates enables discussion and comparison of the codecs in more practical terms.

Peace...

I think he's questioning your defense of MPEG-2 rather than comparing anything to HD-DVD, which I don't think you've done.

Peace...

WHAT?? Are you saying the rate is lower than D-Theater and not better than some HD on TV??

Benefit of the doubt, MPEG-2 on next-gen DVD will be VBR vs CBR for OTA HD. Thus it's possible to at least match (possibly exceed) OTA PQ at similar bitrates.

Of course if they only look like OTA, I won't be standing in line for either.