MPEG-4 (ISO/IEC 14496) is
an ISO/IEC standard developed by MPEG (Moving Picture Experts Group), the
committee that also developed the MPEG-1 and MPEG-2 standards. MPEG-4 is the result of a continuing international effort
involving hundreds of researchers and engineers from all over the world,
following on from the development of MPEG-2.
MPEG-4 Version 1 became an
International Standard in the first months of 1999. The fully backward compatible extensions under the title of
MPEG-4 Version 2 were frozen at the end of 1999 and became an ISO standard early
in 2000. MPEG-4 was the basis of
the failed DivX disc and player system, but has lived on as a standard for
Thus far, MPEG-4 has no
associated standard encryption/copy protection algorithm.
So, MPEG-4 files that are circulating on the Internet are unencrypted.
The studios are very concerned about the spread of MPEG-4 and have
branded it as “the standard used by video pirates”.
The main reason for the
growing popularity of MPEG-4 is its efficiency. It provides much better video compression than MPEG-2.
After all, it represents an additional 4 or 5 years of R&D beyond
MPEG-2. In a sense, the Moving
Picture Experts Group saw the state of the standard in 1995 as “work in
progress”, but had to freeze it as it then stood, and publish it (as MPEG-2)
so that it could be used for the launch of the DVD.
The bit-rate used for
MPEG-4 can be selected based on the required quality/bit-rate trade-off.
It can operate satisfactorily (obviously at low resolutions) as low as
5kbps and has been used at over 1Gbps in certain applications.
As implemented in most PC software applications (like Roxio), it produces
files that are slightly smaller than MPEG-1 (VCD) files, but with much better
It is unclear exactly how
much more efficient MPEG-4 is than MPEG-2 in compression given similar
amounts of available processing power on which to run the encoding
operation. Standards like MPEG-2
and MPEG-4 are “asymmetrical” in that the way the files are played back
follows fixed rules but the way that the files are generated uses a
“toolkit” of compression components such that the more processing power and
processing time that is available when the compressed file is created, the
better the quality of the video at playback time.
The MPEG-2 files on DVDs have been generated using considerable amounts
of processing power and time. By
contrast, when you generate an MPEG-2 file on a PC, only a limited subset of the
compression toolkit is used (otherwise “rendering” a video would take all
day!). Most of the MPEG-4 files in
circulation have been generated using PCs, rather than professional-quality
implementations of MPEG-4 running on powerful servers.
A fair comparison of MPEG-2 versus MPEG-4 would need a “full”
implementation of MPEG-4, allowed enough processing time to “run to
completion” in same the way that the distributors of DVDs run the MPEG-2
algorithm. No such side-by-side
comparison has been published to date. Nevertheless,
a comparison of MPEG-2 and MPEG-4 versions of the same raw material, generated
on the same PC using implementations of the algorithms that run for roughly the
same number of minutes, shows that MPEG-4 is clearly superior in terms of
compression efficiency and video quality.
Various articles and websites quote figures for MPEG-4 compression efficiency ranging from 3 to 10 times as efficient as MPEG-2. One report claims that DVD quality can be obtained at an average bit-rate of only 0.88 Mbps, making it over five times better than MPEG-2. The www.divx.com website claims that it gives compression “up to ten times” that of MPEG-2. (This probably assumes ideal conditions with a powerful encoding engine.) Whether these figures include audio or not is unclear. This makes a big difference. On DVDs, with multiple audio channels in use, up to 1.152 Mbps may be consumed by audio (i.e. almost as much bandwidth as a normal, uncompressed audio signal on a normal CD - see footnote*).
So, if the video part of the signal is, say, 4.3 times more compressed
than with MPEG-2 (0.88Mbps average bit-rate versus 3.8Mbps average bit-rate),
the resulting aggregate average bit-rate for MPEG-4 would be:
0.88Mbps (video) + 1.152Mps (audio) + 0.40Mbps (subtitles) = 2.072Mpbs
In other words, the video
may consume less bandwidth than the audio!
If the audio capacity is
restricted to just one full-quality stereo channel and one lower-quality channel
(e.g. for a Director’s commentary), then the resulting bit stream could be
recorded on a normal CD. Or, to put
it another way, if MPEG-4 had been developed in time for the launch of the VCD,
the VCD could have achieved present-day DVD quality (although it would still
have needed two discs for most full-length movies). Clearly, MPEG-4 is the standard that DVDs would have
used if DVDs had been launched in 2001 instead of 1997. Use of MPEG-4 would have obviated the need to develop, in the
short term at least, dual-layer disc technology.
It seems likely that the
change to MPEG-4, or its successor, will be made with the next generation of
“blue laser” HDTV-DVDs.
In the meantime the Moving
Picture Experts Group seems to not want to annoy the studios that sponsor its
work, so they do not publish the results of any comparative tests of MPEG-4
versus MPEG-2. This is why only
“unofficial” claims about the efficiency of MPEG-4 are available in
publications and on the Internet.
Up until 2002 MPEG-4 was used mainly by video enthusiasts - as the preferred method of saving personal videos on PC hard drives and exchanging video material on the Internet. Bill Gates seems to have sided with the studios in their paranoid response to MPEG-4. In 2002 Microsoft effectively withdrew support for MPEG-4 from Windows Media Player. Media Player connects to Microsoft's website whenever it encounters a file that uses a codec that is not on the user's machine. After Microsoft removed their original MPEG-4 codec from their site, anyone who had not already acquired the codec on their PC would get an error message, when trying to play an MPEG-4 AVI file, saying that no codec was available to play the file.
What they did next was to write a new "improved" codec for "MPEG-4 V3". In fact, MPEG has never issued a Version 3 of MPEG-4. "MPEG-4 V3" is a fictitious designation for a Microsoft interpretation of MPEG-4. They gave the new codec exactly the same registry name and DLL name as the old MPEG-4 codec (VIDC.MPG4, mpg4c32.dll), thus maximizing the chances that users will accidentally overwrite the old mpg4c32.dll when doing any kind of upgrade to Media Player or installing a codec "package". What "MPEG-4 V3" does is to prevent you from playing AVI files. It will only work with Microsoft ASF files. (See details of the Microsoft formats in the next-but-one section.) In effect, Microsoft is trying to hijack MPEG-4 and morph it into ASF, so that all the "copyright policing" that is built into the recent versions of Media Player can be brought into effect.
Microsoft’s website states its position on MPEG-4 as follows:
“While Microsoft continues to support the MPEG-4 standardization process, it is moving forward with the development of audio
and video technologies that deliver superior quality and an end-to-end streaming solution for Microsoft customers.”
setting aside Microsoft's antagonism to un-Microsoftized MPEG-4, there have been
signs of wider commercial adoption of MPEG-4. Steve Jobs has emerged as a strong supporter of MPEG-4.
Apple made a big show of supporting MPEG-4 in QuickTime 6.0, and Steve
Jobs made a point, during the product launch, of deriding Microsoft for shunning
MPEG-4. More importantly,
manufacturers (notably Panasonic, JVC, and Toshiba – though of course not
Sony) have started to incorporate MPEG-4 into products.
For example, some digital cameras with a “short video” feature (the
ability to use the camera as a stand-in for a camcorder for a minute or two of
video) now take advantage of MPEG-4’s performance in order to get as much
video as possible onto the camera’s memory card or disc.
Also, some camcorder manufacturers have started to include in their
products a feature to save a section of video on a memory card as an “MPEG-4
MPEG-4 is being continuously refined (by MPEG) by the addition of new coding options. However, this is being done under strict backwards-compatibility rules. MPEG-4, with the most recent extension to MPEG-4 known as MPEG-4 Part 10, has been registered as an ITU-T standard called H.264. This has added to confusion about the status of MPEG-4, leading to magazine articles saying that “H.264 has leapfrogged MPEG-4”. It seems likely that further improvements will continue to be made to MPEG-4 before an agreed set of extensions is frozen for commercial use (e.g. as a basis of blue-laser high-definition DVDs). Alternatively, the Motion Picture Experts Group may decide to give the resulting collection of definitions a new name (such as MPEG-5) in order to reduce confusion about which parts of the standards documentation are to be used as the standard for commercial deployment.
[Also: See My letter to Widescreen Review about MPEG-4]
*Footnote: the bandwidth of a CD (that is, the net rate at which data is read off a CD, after error-control and framing bits have been stripped off, with the disc spinning with the "normal" linear velocity relative to the lens) is about 1.4 Mbps. The corresponding bandwidth of a DVD is about 9.8 Mbps when the disc is spinning at "normal maximum" speed. In this context, "normal" means the speed used when playing an audio CD or VCD, as opposed to writing a CD or reading data off a CD-ROM, which may be at many times the normal speed, for example, "x24". For a DVD, "maximum normal" similarly relates to playing a video DVD; but this is the maximum allowed speed, used only in parts of a film where there is a great deal of motion. For slower-moving scenes, the read-rate is slowed considerably: on average, across a whole disc, the read-rate is typically around 4 Mbps. If a film were coded in MPEG-2 "flat-out", at 9.8 Mbps, a full-length film would not fit on a single-layer DVD. The playing time of a single-layer DVD at the normal maximum speed is about the same as the playing time of an audio CD - around 74 minutes, since the total capacity of a single DVD layer is about 7 times that of a CD, and the normal maximum read-rate of a DVD is also 7 times that of a CD (9.8/1.4 = 7). In order to fit a full-length film onto a DVD, along with all the extras, the quality of the video is decreased during the MPEG-2 coding process to bring the average bandwidth down to around 4 Mbps. (If you want to see how great MPEG-2 looks when run flat-out at close to 9.8 Mbps, get a copy of Animusic 1.)