Thursday, August 01, 2002 10:09:59 AM
"Everyone seems to be enamored with MPEG-4. But other technologies can squeeze your video too."
Crush story
Paul G Schreier, Contributing Editor -- 8/1/2002
CommVerge
Digital video holds huge possibilities as a convergence application. Unfortunately, the adjective "huge" can also be applied to raw digital video files. Over the years, the ingenious work of compression wizards has resulted in tools that crunch those files down to manageable size. That's why today's consumers can, for instance, store a feature film (plus extras and several soundtracks) on a DVD or view adequate video images over the decidedly video-hostile Web.
But get ready, because compression technologies about to explode on the scene will make it possible to play digital video everywhere, even on PDAs and cell phones. Among these emerging formats, MPEG-4 has created by far the most buzz. However, it has also created a lot of confusion. Though MPEG-4 is promising and is gaining momentum, the standard everybody loves to hype is not the only game in town.
When examining technologies in use today and looking to the future, it's important to separate Internet-based video from non-PC applications. Today, virtually every DVD player or set-top box uses an MPEG-2 decoder. Meanwhile, a large number of chip vendors make MPEG-2 chips, which vary widely in how they implement the scheme, trading off among key parameters such as file size, video quality, and processing power.
Online extras:
A many-splendored thing: An expanded explanation of MPEG-4
One-stop licensing: The MPEG-4 brouhaha
On the PC side, in contrast, the vast majority of the streaming video coming over the Internet today uses proprietary compression schemes. The leading contenders are RealNetworks with its RealVideo format, QuickTime from Apple Computer (long ago made available under Windows as well), and Microsoft with its Windows Media Player, which works with Microsoft's own ASF (advanced streaming format) and AVI (audio/video interleaved) formats.
Both the Microsoft and Real players can work with a variety of formats, and can even download additional codecs to deal with new ones. Often this process is transparent to the end user, who doesn't know or care—as long as the streaming video works.
Such accomplishments notwithstanding, there are limits to what today's codecs can accomplish. Unless you're satisfied with matchbook-sized display windows and stuttering videos, you'll likely agree that we need a better way to move video. And you'll feel even more strongly about the issue if you're among those who hope that video will one day flow to handheld devices over relatively low-bandwidth cellular links.
MPEG-4 has been widely hyped as the standard that's supposed to save us, thanks to stronger compression and nifty features like the ability to send streams of different quality levels out from a single stored file. "MPEG-4 creates the basis for all future developments in digital video," predicts Ben Waggoner, president of digital-video consulting firm Interframe Media and author of the upcoming book, Compression for Great Digital Video: Power Tips, Techniques and Common Sense .
That may very well be correct, but it's still to early to tell; we're only starting to get our first taste of MPEG-4. In June, Apple released a public preview of QuickTime 6, the first generally available MPEG-4 player. So we can soon expect to see individuals and companies making MPEG-4-encoded video available for streaming.
But hang on, you might say, isn't MPEG-4 video already available? After all, many companies have been talking about their MPEG-4 products for some time now. However, until recently, true MPEG-4 video hasn't been available, Waggoner notes.
What about the hugely popular DivX format, which has a large user base and is touted as "MPEG-4-based?" Waggoner has conducted interoperability tests with a large number of players, including some prerelease versions, and found that DivX files won't run on MPEG-4 players.
"Although it leverages some MPEG-4 technology, DivX does not make a true MPEG-4-compatible file," Waggoner says. "It creates an AVI file using MPEG-4 video and MPEG-3 audio." Even the FAQ file at www.divx.com admits the incompatibility: "While we test the DivX Player on a wide variety of third-party MPEG-4 files, we cannot claim full compatibility with all MPEG-4 files due to the wide breadth of the MPEG-4 standard. (We do claim compatibility with MPEG-4 files created with the DivX codec.)"
“MPEG-4 creates the basis for all future developments in digital video.”
—Ben Waggoner, Interframe Media
Even so, DivX has become the format of choice for sharing lengthy video files, such as feature films, on the Internet. In fact, it has reached a status almost equivalent to that of MP3 for audio. One reason is that its compression can pack a typical feature film onto a single CD-R disk. Another plus is that it's free for personal usage. DivX estimates that 50 million users worldwide have its software on their systems.
PC-based video players that work with DivX files are readily available, but unless your graphics card has a TV port, you're stuck on the small screen. The only alternative is to convert the movie to a VCD or SVCD format, burn a CD-R, and play that disc on a conventional DVD player (although hackers have reportedly discovered how to modify Microsoft XBox consoles to play DivX videos). In an effort to expand its core technology into other areas, such as net appliances, DivX is working with Texas Instruments to develop a DivX-enabled DSP that supports DivX playback at full video frame rate and frame size.
Mixed vibrations
Returning to MPEG-4, even though true MPEG-4 video content is just starting to appear, there's plenty of buzz about that codec—and just as many questions. Will its additions and improvements be enough to attract attention from developers and users?
"In the profiles people are working with today, MPEG-4 doesn't add any new functionality to the video world," Waggoner says, referring to the standard's usage profiles, which define subsets of features tailored for various applications (see the sidebar, "A many-splendored thing"). "Later, though, developers will be able to add interactivity to their content. Next year you can expect to see lots of interesting functionality emerge."
A few more specifics come from Didier LeGall, vice president for home media products at LSI Logic. "The MPEG-4 Simple Profile isn't really a breakthrough compared to MPEG-2," he says. "Instead, it's roughly equivalent. A more advanced solution is coming in other formats. For instance, the Advanced Simple Profile offers at least a 15 percent compression advantage compared with MPEG-2."
If 15 percent doesn't sound overwhelming, don't despair. There's another promising format out there, which isn't currently part of MPEG-4 but will likely become so down the road. "The technology of the future is based on the work of the H.26L project of the ITU [International Telecommunications Union]," LeGall says. "This breakthrough achieves a 45 to 50 percent gain in compression efficiency compared with MPEG-2."
This is a case where the technology and the associated groups are evolving so rapidly that it can be difficult to tell the players without a scorecard. H.26L began its life in 1977 under the auspices of the ITU-T Video Coding Experts Group. Today, it appears to be far superior to MPEG-4 as presently defined. Most of the performance gains come from improved motion estimation.
“To have a format broadly accepted, you must freeze the format at some time. And once MPEG-4 Part 10 becomes widely available, the competitive advantage of proprietary codecs will go away.”
—Rob Koenen, MPEG-4 Industry Forum
Studies by UBVideo, which is working to implement this technology on a Texas Instruments DSP, show that compared with MPEG-4's Simple Profile, H.26L permits an average reduction of 50 percent in bit rate for a similar degree of encoder optimization. At the same time, it offers consistently high video quality at all bit rates. Furthermore, it can operate in a low-delay mode to adapt to real-time communications applications and is also error-resilient, with the ability to deal with packet loss and bit errors.
The scheme shows such promise that the ITU has joined with another standards body, ISO/IEC (International Organization for Standards/International Electrotechnical Commission), to coordinate activity on video coding. The new working party is called the Joint Video Team (JVT), and its work will likely produce two key results. The first is a new ITU-T Recommendation, which, if it follows standard naming conventions, will become H.264. The second is an addition to the MPEG-4 standard, likely to carry the name MPEG-4 Part 10. For now, H.26L, H.264, and MPEG-4 Part 10 all refer to the same work.
As for a timeline, Rob Koenen, president of the MPEG-4 Industry Forum (M4IF), predicts that the standard, including H.26L, should be frozen by the end of this year. We should see software shortly thereafter, with hardware following later, he adds.
However, warns Doug McIntyre, CEO of compression-technology supplier On2 Technologies, "H.26L won't be commercially viable for a couple of years. Yes, it will be possible to compress video to a high degree, but the rate today is a frame per minute. It's good technology, but it has not yet been optimized and it will need a couple of years before it sees productization."
LSI Logic's LeGall, agrees—somewhat. "Yes, some vendors of proprietary codecs point out that H.26L runs very slowly on PCs and it's impractical today," he says. "But in two years it will work in real time and compete strongly with those suppliers."
Bucking the standard
While considerable work is taking place on the MPEG-4 front, proprietary codecs continue to attract considerable usage. "Proprietary codecs are still somewhat better," admits the M4IF's Koenen. "Their developers can release new versions every half year, taking advantage of the latest research. But to have a format broadly accepted, you must freeze the format at some time. Consider the example of MP3 for audio, whose adoption has been great for the overall market. Once MPEG-4 Part 10 becomes widely available, the competitive advantage of proprietary codecs will go away."
“Every two weeks there’s a press release on a new breakthrough in a proprietary codec—or is it a breakthrough in marketing only?”
—Didier LeGall, LSI Logic
LeGall echos that sentiment. "Every two weeks there's a press release on a new breakthrough in a proprietary codec," he says. "Or is it a breakthrough in marketing only? You've got an international standard, and then Microsoft, and you want to compete against that pair with a proprietary codec? Good luck."
LeGall's mention of Microsoft brings us to Corona, the code name for an enhanced media toolset and codec slated for release by the end of this year. Unsurprisingly—considering Microsoft's history—Corona won't play MPEG-4 files and instead will use proprietary technology. Microsoft is making some aggressive claims about Corona's video quality and capabilities, many of which have yet to be borne out in practice. Nonetheless, the technology can't be ignored. "It looks very promising," Waggoner says, "and this year I believe it should perform better than MPEG-4."
In addition, most industry observers believe that Microsoft will ultimately incorporate some level of support for MPEG-4 when the company can no longer ignore the large amount of video content that the standardized codec will attract.
That's a common tack for vendors of proprietary codecs to take. "We all believe that our codecs are far superior to MPEG-4, but we must also ensure interoperability," says Dave Cotter, group products manager in the media-systems division at RealNetworks. "How do we do so? By creating standards in the delivery protocols. In fact, we at RealNetworks agree that the delivery system is the future; design engineers don't really have to worry where the codec train is going."
RealNetworks and Microsoft are not going to go away, says On2's McIntyre. "So any codecs that interoperate with these players have a better chance of survival," he says. "I'd be betting on a distribution system—not on which codec works in it. Those two companies will always try to push the best codecs or use their considerable engineering staffs to write good ones."
Given these comments, it's interesting that On2 hasn't made its own latest codec technology compatible with Windows Media Player. "Support for Apple and RealNetworks is good enough," McIntyre says. "As of yet, there's no commercial reason to move to Windows Media Player. We believe that 90 percent of all PCs have both RealPlayer and Windows Media Player installed, and as long as we run on one of them, I'm fine."
In the hand
The idea of being able to upgrade or enhance a media player with multiple codecs is easy to envision in the PC world. But the concept doesn't transfer quite as well to closed devices, where codec code is burned into memory. Certainly, MPEG-4 won't be limited to streaming applications on PCs any more than is MPEG-2, which today is incorporated into virtually every DVD player.
“I’d be betting on a distribution system—not on which codec works in it.”
—Doug McIntyre, On2 Technologies
"The key that will drive this market consists of the MPEG-4 chips being developed," comments Kent Libbey, vice president of marketing at iVast, a vendor of MPEG-4 software. "Chips made the MPEG-2 market what it is today, because DVDs, PVRs [personal video recorders], and satellite boxes all use them." iVast supplies the codec technology behind e-Box, an effort among seven companies, including Sharp and Pioneer, to bring MPEG-4 to set-top boxes. Cable provider Comcast is also collaborating and will be the first to test the system on its release next year.
A market shakeout in the number of codecs, led by MPEG-4, will also help chip vendors. "We don't want to deploy 50 codecs," says LSI Logic's LeGall. "We do about a dozen now, and each must be a value proposition." The company's DoMiNo architecture supports a range of codecs, including DV (for camcorders), MPEG-1, MPEG-2, and MPEG-4 Simple Profile. In the future the family will also support the Advanced Simple Profile as well as Part 10 and even Microsoft's Corona. "We're not sure that just one format will be a winner, and we look at solutions that accommodate multiple formats," LeGall says.
That sounds like good advice for any player eyeing the digital-video space.
Author information
Contributing Editor Paul G Schreier (pgschreier@amitechmarketing.com) is waiting for the compression algorithm that can efficiently compact weeks of research into a 3000-word feature article.
A many-splendored thing
If you've seen the term MPEG-4 thrown around in seemingly incompatible contexts, don't be surprised. Rather than a single, universally applicable standard, MPEG-4 is best understood as a multimedia framework that addresses the needs of a wide variety of applications—including but not limited to studio editing, interactive broadcasts, Internet streaming, and playback on wireless devices.
Today, MPEG-4 is broken down into eight separate major parts (ISO/IEC-14496-x), two of which address visual and audio functions. However, because the standard is object based, it extends far beyond those two areas to support interactive multimedia, which will play an important role in the future.
You may hear that MPEG-4 and Apple's QuickTime are the same, but that's not really the case. The MPEG-4 group selected the QuickTime file format to start with, so MPEG-4 software does share a deep level of compatibility with QuickTime. However, the standard has evolved in scope and functionality, so the two are no longer directly compatible.
While the MPEG-4 standard includes hundreds of features, no application will need all of them. Depending on the target application—be it digital cinema or cell-phone video—only a subset of capabilities will be invoked. To help developers of encoders and decoders work under some common assumptions, the standards group has defined a number of profiles (which define features from a qualitative aspect) and levels (which nail down specific performance requirements for these features).
The formal specification lists 19 different Visual Profiles, "only a few of which will actually be used," says Rob Koenen, president of the MPEG-4 Industry Forum.
Here are some details on the profiles you're most likely to encounter in the wild:
Simple Profile: Provides a technology baseline upon which the other profiles build, and is suited for use on mobile or other low-power devices.
Advanced Simple Profile: Adds enhancements for better visual quality, such as global-motion compensation and quarter-pixel motion estimation (Simple uses half-pixel motion estimation).
Simple Scalable Profile: Allows a server to modify the bit rate depending on demands on image quality, frame rate, or resolution. Useful for applications that provide services at more than one level of quality due to bandwidth or decoder limitations.
Fine Granularity Scalable (FGS): Will allow for extremely high quality video.
Core Profile: Adds support for coding of arbitrary-shaped and temporally scalable objects. Useful for applications such as those providing relatively simple content-interactivity (Internet multimedia applications).
Core Scalable Profile: Adds variable frame rates and resolutions.
Main Profile: Supports interlaced video and appears to be a candidate to support interactive HDTV. To the Core Visual Profile, adds support for coding of interlaced, semi-transparent, and sprite objects. Useful for interactive and entertainment-quality broadcast and DVD applications.
The profiles spell out some generic functions and requirements, but also leave plenty of room for interpretation, such as how intensively the underlying technology will operate. But developers need to nail down more specifics for implementation. This is where the levels come into play.
In any given profile, each level dictates a least-common-denominator with regard to key operating specs, such as the maximum frame rate, data rate, resolution, and number of objects. With levels, a developer can make certain assumptions about the ultimate system.
For example, within the Simple Profile, Level 0 targets the wireless industry, where screen sizes are small and processing power is low. The level allows a maximum frame rate of 15 frames/sec, limits luminance pixel resolutions, supports 99 macroblocks (16-by-16-pixel areas), and caps overall screen size at 176 by 144 pixels. Other levels don't dictate a max image size, but they can specify the number of macroblocks. For example, Level 1 allows 99 macroblocks that can be configured as 1584 by 16, 176 by 144, or any other rectangular size divisible in each direction.
Despite these commonalities, MPEG-4 products designed to the same profile/level can vary dramatically in how they trade off speed and visual quality. For instance, the standard does not specify how to transport bits over various networks. "The standard defines a bitstream," explains Ben Waggoner, president of digital-video consulting firm Interframe Media. "But there's a tremendous amount of room for innovation at either end of the stream. At the encoder end, you can trade off quality versus speed. On the playback side, the developers can look for ways to get smoother playback. When MPEG-2 first came out, we needed 4 Mbits/sec to achieve DVD quality, but that has since improved to 2 Mbits/sec for a good picture. We'll likewise see MPEG-4 improve as the encoders get better."
And if all this isn't complicated enough, note that the ISMA (Internet Streaming Media Alliance, www.isma.tv) has also defined a common set of requirements and functions—and has unfortunately chosen to use the "profile" nomenclature. For instance, the ISMA's Profile 1 is the equivalent of the Advanced Simple Profile at Level 3 combined with the Audio Profile at Level 2.
This is an important combination, Waggoner says. "The ISMA Profile 1 will become what everyone means by MPEG-4 this year," he says, adding that Profile 1 isn't as good in quality as proprietary codecs and that companies won't migrate toward it for that reason. Instead, they'll like the broad interoperability it affords. "Given MPEG-4, I don't see that somebody like AOL Time-Warner will want to get tied down with a proprietary codec," he comments.
Streaming at windmills?
One hears a lot of hype about streaming video to cell phones and other wireless handhelds, but it's going to be a while before that application becomes a reality. In the meantime, the PC streaming market offers a potentially sobering lesson. To wit: It's not clear whether anyone can turn a profit streaming video to PCs.
"There are very few business models for streaming video that make any money," says Doug McIntyre, CEO of compression-technology supplier On2 Technologies. "Perhaps the only one is porn. Will people watch long-form video on PCs? I think not. In fact, the amount of streaming video on the Web, particularly on multimedia sites, is contracting. People are retreating from this side of the market because there's no way to make money. In the future, I also see decreasing use of video players because consumers will use PCs as a gateway to move video data to a set-top box or PVR [personal video recorder] for more comfortable viewing."
People are more likely to use PCs for specific applications, such as videoconferencing and distance learning, he continues. "For those applications you'll want interaction and a more sophisticated GUI," he says. "Those programs will be far more than simple video players."
Not surprisingly, a representative from RealNetworks has a different viewpoint. Dave Cotter, group products manager in the company's media-systems division, argues that it will be possible to make money with streaming video. "It's not that the model is flawed, but that the technology has been too expensive," he asserts. Higher-quality content that will attract users is becoming available. "In addition, bandwidth costs are dropping, in conjunction with our compression, making streaming video more cost-effective," he says.
One-stop licensing
MPEG-4 is a complex scheme that exists thanks to the work of many companies, which hold patents for various portions of the underlying technology. Rather than make users of the technology approach each company individually for a license, a company called MPEG LA offers a one-stop licensing service. It collects fees on behalf of the patent owners and distributes the proceeds. MPEG LA has successfully shown how this approach works by creating a patent portfolio for the MPEG-2 video compression standard, which it began licensing in 1997.
In January of this year, the company announced its intentions to do the same with portions of the MPEG-4 standard, specifically patents essential to MPEG-4 Visual (Simple and Core Profiles), which involves patents owned by 20 companies. MPEG LA also disclosed a proposed fee structure, under which licensees would pay 25 cents each for MPEG-4 decoders and encoders, with fees capped at $1 million a year for each licensee. The company also suggested charging a per-minute rate with no cap.
These terms created an immediate uproar, with some companies believing they would make MPEG-4 nonviable and open the door for proprietary formats. Other companies were more sanguine. "We took MPEG LA's initial proposal as an opening round of negotiations, while others took it as a fait accompli and overreacted," says Kent Libby, vice president of marketing for iVast. "We analyzed it and reported our findings to the company, as did other responsible parties. It's a living scheme, and things will work out fine. The same process happened with MPEG-2."
True to this prediction, MPEG LA has softened its stance. In mid-July, the company announced the terms for the licenses it will officially roll out in September. The revised terms feature "reasonable" annual limitations on certain royalties, royalty options that require no royalty reports, and threshold levels below which some use-based royalties would not be charged, according to MPEG LA.
"This sounds the starting bell for the whole broadcast and multimedia industry to start releasing MPEG-4 products and services," said Rob Koenen, president of the MPEG-4 Industry Forum, in the press release announcing the terms. "The licenses are the long-expected prerequisite for MPEG-4 being fully accepted and deployed."
But the fact that MPEG LA has revised its license fee structure hasn't removed all objections to the scheme. In fact, Doug McIntyre, CEO of On2 Technologies, a developer of a proprietary codec, argues that there's an even bigger problem. He asserts that MPEG LA is running an illegal patent pool that has not been approved or sanctioned.
"While the MPEG-2 group did get provisional approval for their activities, to my knowledge the MPEG-4 group has not even applied for approval," he says. "It's simply illegal to set pricing in this fashion. It creates a barrier-to-entry for others. We've filed complaints with all 50 state attorneys general, as well as the US Justice Department Antitrust Division and the Federal Trade Commission."
Meanwhile, you may wonder how companies that are talking up MPEG-4 codecs can do so without having signed licensing agreements. "Technically," notes Larry Horn, MPEG LA's vice president of licensing and business development, "they're violating the patents, because the license covers MPEG-4 video in any form or shape. All uses are liable to the patent holders." Meanwhile, MPEG LA looks upon the use of MPEG-4 as a vote of confidence in the technology and the licensing scheme.
"We hope to have a product that allows users to get whatever they need from one place," he concludes.
The main text of this article indicates that the future may lie with H.26L (Part 10), which, Horn notes, is completely divorced from MPEG LA's efforts.
Also, there may be questions about codecs that are billed as "MPEG-4-based" yet don't play on MPEG-4 decoders. "If they use MPEG-4, they will need our license," he says. "But otherwise you have to take a closer look at what 'MPEG-4-based' means. They might need a license for some of the underlying patents from individual patent holders."
http://www.e-insite.net/commvergemag/index.asp?layout=article&articleid=CA236131&pubdate=8/1....
culater
Crush story
Paul G Schreier, Contributing Editor -- 8/1/2002
CommVerge
Digital video holds huge possibilities as a convergence application. Unfortunately, the adjective "huge" can also be applied to raw digital video files. Over the years, the ingenious work of compression wizards has resulted in tools that crunch those files down to manageable size. That's why today's consumers can, for instance, store a feature film (plus extras and several soundtracks) on a DVD or view adequate video images over the decidedly video-hostile Web.
But get ready, because compression technologies about to explode on the scene will make it possible to play digital video everywhere, even on PDAs and cell phones. Among these emerging formats, MPEG-4 has created by far the most buzz. However, it has also created a lot of confusion. Though MPEG-4 is promising and is gaining momentum, the standard everybody loves to hype is not the only game in town.
When examining technologies in use today and looking to the future, it's important to separate Internet-based video from non-PC applications. Today, virtually every DVD player or set-top box uses an MPEG-2 decoder. Meanwhile, a large number of chip vendors make MPEG-2 chips, which vary widely in how they implement the scheme, trading off among key parameters such as file size, video quality, and processing power.
Online extras:
A many-splendored thing: An expanded explanation of MPEG-4
One-stop licensing: The MPEG-4 brouhaha
On the PC side, in contrast, the vast majority of the streaming video coming over the Internet today uses proprietary compression schemes. The leading contenders are RealNetworks with its RealVideo format, QuickTime from Apple Computer (long ago made available under Windows as well), and Microsoft with its Windows Media Player, which works with Microsoft's own ASF (advanced streaming format) and AVI (audio/video interleaved) formats.
Both the Microsoft and Real players can work with a variety of formats, and can even download additional codecs to deal with new ones. Often this process is transparent to the end user, who doesn't know or care—as long as the streaming video works.
Such accomplishments notwithstanding, there are limits to what today's codecs can accomplish. Unless you're satisfied with matchbook-sized display windows and stuttering videos, you'll likely agree that we need a better way to move video. And you'll feel even more strongly about the issue if you're among those who hope that video will one day flow to handheld devices over relatively low-bandwidth cellular links.
MPEG-4 has been widely hyped as the standard that's supposed to save us, thanks to stronger compression and nifty features like the ability to send streams of different quality levels out from a single stored file. "MPEG-4 creates the basis for all future developments in digital video," predicts Ben Waggoner, president of digital-video consulting firm Interframe Media and author of the upcoming book, Compression for Great Digital Video: Power Tips, Techniques and Common Sense .
That may very well be correct, but it's still to early to tell; we're only starting to get our first taste of MPEG-4. In June, Apple released a public preview of QuickTime 6, the first generally available MPEG-4 player. So we can soon expect to see individuals and companies making MPEG-4-encoded video available for streaming.
But hang on, you might say, isn't MPEG-4 video already available? After all, many companies have been talking about their MPEG-4 products for some time now. However, until recently, true MPEG-4 video hasn't been available, Waggoner notes.
What about the hugely popular DivX format, which has a large user base and is touted as "MPEG-4-based?" Waggoner has conducted interoperability tests with a large number of players, including some prerelease versions, and found that DivX files won't run on MPEG-4 players.
"Although it leverages some MPEG-4 technology, DivX does not make a true MPEG-4-compatible file," Waggoner says. "It creates an AVI file using MPEG-4 video and MPEG-3 audio." Even the FAQ file at www.divx.com admits the incompatibility: "While we test the DivX Player on a wide variety of third-party MPEG-4 files, we cannot claim full compatibility with all MPEG-4 files due to the wide breadth of the MPEG-4 standard. (We do claim compatibility with MPEG-4 files created with the DivX codec.)"
“MPEG-4 creates the basis for all future developments in digital video.”
—Ben Waggoner, Interframe Media
Even so, DivX has become the format of choice for sharing lengthy video files, such as feature films, on the Internet. In fact, it has reached a status almost equivalent to that of MP3 for audio. One reason is that its compression can pack a typical feature film onto a single CD-R disk. Another plus is that it's free for personal usage. DivX estimates that 50 million users worldwide have its software on their systems.
PC-based video players that work with DivX files are readily available, but unless your graphics card has a TV port, you're stuck on the small screen. The only alternative is to convert the movie to a VCD or SVCD format, burn a CD-R, and play that disc on a conventional DVD player (although hackers have reportedly discovered how to modify Microsoft XBox consoles to play DivX videos). In an effort to expand its core technology into other areas, such as net appliances, DivX is working with Texas Instruments to develop a DivX-enabled DSP that supports DivX playback at full video frame rate and frame size.
Mixed vibrations
Returning to MPEG-4, even though true MPEG-4 video content is just starting to appear, there's plenty of buzz about that codec—and just as many questions. Will its additions and improvements be enough to attract attention from developers and users?
"In the profiles people are working with today, MPEG-4 doesn't add any new functionality to the video world," Waggoner says, referring to the standard's usage profiles, which define subsets of features tailored for various applications (see the sidebar, "A many-splendored thing"). "Later, though, developers will be able to add interactivity to their content. Next year you can expect to see lots of interesting functionality emerge."
A few more specifics come from Didier LeGall, vice president for home media products at LSI Logic. "The MPEG-4 Simple Profile isn't really a breakthrough compared to MPEG-2," he says. "Instead, it's roughly equivalent. A more advanced solution is coming in other formats. For instance, the Advanced Simple Profile offers at least a 15 percent compression advantage compared with MPEG-2."
If 15 percent doesn't sound overwhelming, don't despair. There's another promising format out there, which isn't currently part of MPEG-4 but will likely become so down the road. "The technology of the future is based on the work of the H.26L project of the ITU [International Telecommunications Union]," LeGall says. "This breakthrough achieves a 45 to 50 percent gain in compression efficiency compared with MPEG-2."
This is a case where the technology and the associated groups are evolving so rapidly that it can be difficult to tell the players without a scorecard. H.26L began its life in 1977 under the auspices of the ITU-T Video Coding Experts Group. Today, it appears to be far superior to MPEG-4 as presently defined. Most of the performance gains come from improved motion estimation.
“To have a format broadly accepted, you must freeze the format at some time. And once MPEG-4 Part 10 becomes widely available, the competitive advantage of proprietary codecs will go away.”
—Rob Koenen, MPEG-4 Industry Forum
Studies by UBVideo, which is working to implement this technology on a Texas Instruments DSP, show that compared with MPEG-4's Simple Profile, H.26L permits an average reduction of 50 percent in bit rate for a similar degree of encoder optimization. At the same time, it offers consistently high video quality at all bit rates. Furthermore, it can operate in a low-delay mode to adapt to real-time communications applications and is also error-resilient, with the ability to deal with packet loss and bit errors.
The scheme shows such promise that the ITU has joined with another standards body, ISO/IEC (International Organization for Standards/International Electrotechnical Commission), to coordinate activity on video coding. The new working party is called the Joint Video Team (JVT), and its work will likely produce two key results. The first is a new ITU-T Recommendation, which, if it follows standard naming conventions, will become H.264. The second is an addition to the MPEG-4 standard, likely to carry the name MPEG-4 Part 10. For now, H.26L, H.264, and MPEG-4 Part 10 all refer to the same work.
As for a timeline, Rob Koenen, president of the MPEG-4 Industry Forum (M4IF), predicts that the standard, including H.26L, should be frozen by the end of this year. We should see software shortly thereafter, with hardware following later, he adds.
However, warns Doug McIntyre, CEO of compression-technology supplier On2 Technologies, "H.26L won't be commercially viable for a couple of years. Yes, it will be possible to compress video to a high degree, but the rate today is a frame per minute. It's good technology, but it has not yet been optimized and it will need a couple of years before it sees productization."
LSI Logic's LeGall, agrees—somewhat. "Yes, some vendors of proprietary codecs point out that H.26L runs very slowly on PCs and it's impractical today," he says. "But in two years it will work in real time and compete strongly with those suppliers."
Bucking the standard
While considerable work is taking place on the MPEG-4 front, proprietary codecs continue to attract considerable usage. "Proprietary codecs are still somewhat better," admits the M4IF's Koenen. "Their developers can release new versions every half year, taking advantage of the latest research. But to have a format broadly accepted, you must freeze the format at some time. Consider the example of MP3 for audio, whose adoption has been great for the overall market. Once MPEG-4 Part 10 becomes widely available, the competitive advantage of proprietary codecs will go away."
“Every two weeks there’s a press release on a new breakthrough in a proprietary codec—or is it a breakthrough in marketing only?”
—Didier LeGall, LSI Logic
LeGall echos that sentiment. "Every two weeks there's a press release on a new breakthrough in a proprietary codec," he says. "Or is it a breakthrough in marketing only? You've got an international standard, and then Microsoft, and you want to compete against that pair with a proprietary codec? Good luck."
LeGall's mention of Microsoft brings us to Corona, the code name for an enhanced media toolset and codec slated for release by the end of this year. Unsurprisingly—considering Microsoft's history—Corona won't play MPEG-4 files and instead will use proprietary technology. Microsoft is making some aggressive claims about Corona's video quality and capabilities, many of which have yet to be borne out in practice. Nonetheless, the technology can't be ignored. "It looks very promising," Waggoner says, "and this year I believe it should perform better than MPEG-4."
In addition, most industry observers believe that Microsoft will ultimately incorporate some level of support for MPEG-4 when the company can no longer ignore the large amount of video content that the standardized codec will attract.
That's a common tack for vendors of proprietary codecs to take. "We all believe that our codecs are far superior to MPEG-4, but we must also ensure interoperability," says Dave Cotter, group products manager in the media-systems division at RealNetworks. "How do we do so? By creating standards in the delivery protocols. In fact, we at RealNetworks agree that the delivery system is the future; design engineers don't really have to worry where the codec train is going."
RealNetworks and Microsoft are not going to go away, says On2's McIntyre. "So any codecs that interoperate with these players have a better chance of survival," he says. "I'd be betting on a distribution system—not on which codec works in it. Those two companies will always try to push the best codecs or use their considerable engineering staffs to write good ones."
Given these comments, it's interesting that On2 hasn't made its own latest codec technology compatible with Windows Media Player. "Support for Apple and RealNetworks is good enough," McIntyre says. "As of yet, there's no commercial reason to move to Windows Media Player. We believe that 90 percent of all PCs have both RealPlayer and Windows Media Player installed, and as long as we run on one of them, I'm fine."
In the hand
The idea of being able to upgrade or enhance a media player with multiple codecs is easy to envision in the PC world. But the concept doesn't transfer quite as well to closed devices, where codec code is burned into memory. Certainly, MPEG-4 won't be limited to streaming applications on PCs any more than is MPEG-2, which today is incorporated into virtually every DVD player.
“I’d be betting on a distribution system—not on which codec works in it.”
—Doug McIntyre, On2 Technologies
"The key that will drive this market consists of the MPEG-4 chips being developed," comments Kent Libbey, vice president of marketing at iVast, a vendor of MPEG-4 software. "Chips made the MPEG-2 market what it is today, because DVDs, PVRs [personal video recorders], and satellite boxes all use them." iVast supplies the codec technology behind e-Box, an effort among seven companies, including Sharp and Pioneer, to bring MPEG-4 to set-top boxes. Cable provider Comcast is also collaborating and will be the first to test the system on its release next year.
A market shakeout in the number of codecs, led by MPEG-4, will also help chip vendors. "We don't want to deploy 50 codecs," says LSI Logic's LeGall. "We do about a dozen now, and each must be a value proposition." The company's DoMiNo architecture supports a range of codecs, including DV (for camcorders), MPEG-1, MPEG-2, and MPEG-4 Simple Profile. In the future the family will also support the Advanced Simple Profile as well as Part 10 and even Microsoft's Corona. "We're not sure that just one format will be a winner, and we look at solutions that accommodate multiple formats," LeGall says.
That sounds like good advice for any player eyeing the digital-video space.
Author information
Contributing Editor Paul G Schreier (pgschreier@amitechmarketing.com) is waiting for the compression algorithm that can efficiently compact weeks of research into a 3000-word feature article.
A many-splendored thing
If you've seen the term MPEG-4 thrown around in seemingly incompatible contexts, don't be surprised. Rather than a single, universally applicable standard, MPEG-4 is best understood as a multimedia framework that addresses the needs of a wide variety of applications—including but not limited to studio editing, interactive broadcasts, Internet streaming, and playback on wireless devices.
Today, MPEG-4 is broken down into eight separate major parts (ISO/IEC-14496-x), two of which address visual and audio functions. However, because the standard is object based, it extends far beyond those two areas to support interactive multimedia, which will play an important role in the future.
You may hear that MPEG-4 and Apple's QuickTime are the same, but that's not really the case. The MPEG-4 group selected the QuickTime file format to start with, so MPEG-4 software does share a deep level of compatibility with QuickTime. However, the standard has evolved in scope and functionality, so the two are no longer directly compatible.
While the MPEG-4 standard includes hundreds of features, no application will need all of them. Depending on the target application—be it digital cinema or cell-phone video—only a subset of capabilities will be invoked. To help developers of encoders and decoders work under some common assumptions, the standards group has defined a number of profiles (which define features from a qualitative aspect) and levels (which nail down specific performance requirements for these features).
The formal specification lists 19 different Visual Profiles, "only a few of which will actually be used," says Rob Koenen, president of the MPEG-4 Industry Forum.
Here are some details on the profiles you're most likely to encounter in the wild:
Simple Profile: Provides a technology baseline upon which the other profiles build, and is suited for use on mobile or other low-power devices.
Advanced Simple Profile: Adds enhancements for better visual quality, such as global-motion compensation and quarter-pixel motion estimation (Simple uses half-pixel motion estimation).
Simple Scalable Profile: Allows a server to modify the bit rate depending on demands on image quality, frame rate, or resolution. Useful for applications that provide services at more than one level of quality due to bandwidth or decoder limitations.
Fine Granularity Scalable (FGS): Will allow for extremely high quality video.
Core Profile: Adds support for coding of arbitrary-shaped and temporally scalable objects. Useful for applications such as those providing relatively simple content-interactivity (Internet multimedia applications).
Core Scalable Profile: Adds variable frame rates and resolutions.
Main Profile: Supports interlaced video and appears to be a candidate to support interactive HDTV. To the Core Visual Profile, adds support for coding of interlaced, semi-transparent, and sprite objects. Useful for interactive and entertainment-quality broadcast and DVD applications.
The profiles spell out some generic functions and requirements, but also leave plenty of room for interpretation, such as how intensively the underlying technology will operate. But developers need to nail down more specifics for implementation. This is where the levels come into play.
In any given profile, each level dictates a least-common-denominator with regard to key operating specs, such as the maximum frame rate, data rate, resolution, and number of objects. With levels, a developer can make certain assumptions about the ultimate system.
For example, within the Simple Profile, Level 0 targets the wireless industry, where screen sizes are small and processing power is low. The level allows a maximum frame rate of 15 frames/sec, limits luminance pixel resolutions, supports 99 macroblocks (16-by-16-pixel areas), and caps overall screen size at 176 by 144 pixels. Other levels don't dictate a max image size, but they can specify the number of macroblocks. For example, Level 1 allows 99 macroblocks that can be configured as 1584 by 16, 176 by 144, or any other rectangular size divisible in each direction.
Despite these commonalities, MPEG-4 products designed to the same profile/level can vary dramatically in how they trade off speed and visual quality. For instance, the standard does not specify how to transport bits over various networks. "The standard defines a bitstream," explains Ben Waggoner, president of digital-video consulting firm Interframe Media. "But there's a tremendous amount of room for innovation at either end of the stream. At the encoder end, you can trade off quality versus speed. On the playback side, the developers can look for ways to get smoother playback. When MPEG-2 first came out, we needed 4 Mbits/sec to achieve DVD quality, but that has since improved to 2 Mbits/sec for a good picture. We'll likewise see MPEG-4 improve as the encoders get better."
And if all this isn't complicated enough, note that the ISMA (Internet Streaming Media Alliance, www.isma.tv) has also defined a common set of requirements and functions—and has unfortunately chosen to use the "profile" nomenclature. For instance, the ISMA's Profile 1 is the equivalent of the Advanced Simple Profile at Level 3 combined with the Audio Profile at Level 2.
This is an important combination, Waggoner says. "The ISMA Profile 1 will become what everyone means by MPEG-4 this year," he says, adding that Profile 1 isn't as good in quality as proprietary codecs and that companies won't migrate toward it for that reason. Instead, they'll like the broad interoperability it affords. "Given MPEG-4, I don't see that somebody like AOL Time-Warner will want to get tied down with a proprietary codec," he comments.
Streaming at windmills?
One hears a lot of hype about streaming video to cell phones and other wireless handhelds, but it's going to be a while before that application becomes a reality. In the meantime, the PC streaming market offers a potentially sobering lesson. To wit: It's not clear whether anyone can turn a profit streaming video to PCs.
"There are very few business models for streaming video that make any money," says Doug McIntyre, CEO of compression-technology supplier On2 Technologies. "Perhaps the only one is porn. Will people watch long-form video on PCs? I think not. In fact, the amount of streaming video on the Web, particularly on multimedia sites, is contracting. People are retreating from this side of the market because there's no way to make money. In the future, I also see decreasing use of video players because consumers will use PCs as a gateway to move video data to a set-top box or PVR [personal video recorder] for more comfortable viewing."
People are more likely to use PCs for specific applications, such as videoconferencing and distance learning, he continues. "For those applications you'll want interaction and a more sophisticated GUI," he says. "Those programs will be far more than simple video players."
Not surprisingly, a representative from RealNetworks has a different viewpoint. Dave Cotter, group products manager in the company's media-systems division, argues that it will be possible to make money with streaming video. "It's not that the model is flawed, but that the technology has been too expensive," he asserts. Higher-quality content that will attract users is becoming available. "In addition, bandwidth costs are dropping, in conjunction with our compression, making streaming video more cost-effective," he says.
One-stop licensing
MPEG-4 is a complex scheme that exists thanks to the work of many companies, which hold patents for various portions of the underlying technology. Rather than make users of the technology approach each company individually for a license, a company called MPEG LA offers a one-stop licensing service. It collects fees on behalf of the patent owners and distributes the proceeds. MPEG LA has successfully shown how this approach works by creating a patent portfolio for the MPEG-2 video compression standard, which it began licensing in 1997.
In January of this year, the company announced its intentions to do the same with portions of the MPEG-4 standard, specifically patents essential to MPEG-4 Visual (Simple and Core Profiles), which involves patents owned by 20 companies. MPEG LA also disclosed a proposed fee structure, under which licensees would pay 25 cents each for MPEG-4 decoders and encoders, with fees capped at $1 million a year for each licensee. The company also suggested charging a per-minute rate with no cap.
These terms created an immediate uproar, with some companies believing they would make MPEG-4 nonviable and open the door for proprietary formats. Other companies were more sanguine. "We took MPEG LA's initial proposal as an opening round of negotiations, while others took it as a fait accompli and overreacted," says Kent Libby, vice president of marketing for iVast. "We analyzed it and reported our findings to the company, as did other responsible parties. It's a living scheme, and things will work out fine. The same process happened with MPEG-2."
True to this prediction, MPEG LA has softened its stance. In mid-July, the company announced the terms for the licenses it will officially roll out in September. The revised terms feature "reasonable" annual limitations on certain royalties, royalty options that require no royalty reports, and threshold levels below which some use-based royalties would not be charged, according to MPEG LA.
"This sounds the starting bell for the whole broadcast and multimedia industry to start releasing MPEG-4 products and services," said Rob Koenen, president of the MPEG-4 Industry Forum, in the press release announcing the terms. "The licenses are the long-expected prerequisite for MPEG-4 being fully accepted and deployed."
But the fact that MPEG LA has revised its license fee structure hasn't removed all objections to the scheme. In fact, Doug McIntyre, CEO of On2 Technologies, a developer of a proprietary codec, argues that there's an even bigger problem. He asserts that MPEG LA is running an illegal patent pool that has not been approved or sanctioned.
"While the MPEG-2 group did get provisional approval for their activities, to my knowledge the MPEG-4 group has not even applied for approval," he says. "It's simply illegal to set pricing in this fashion. It creates a barrier-to-entry for others. We've filed complaints with all 50 state attorneys general, as well as the US Justice Department Antitrust Division and the Federal Trade Commission."
Meanwhile, you may wonder how companies that are talking up MPEG-4 codecs can do so without having signed licensing agreements. "Technically," notes Larry Horn, MPEG LA's vice president of licensing and business development, "they're violating the patents, because the license covers MPEG-4 video in any form or shape. All uses are liable to the patent holders." Meanwhile, MPEG LA looks upon the use of MPEG-4 as a vote of confidence in the technology and the licensing scheme.
"We hope to have a product that allows users to get whatever they need from one place," he concludes.
The main text of this article indicates that the future may lie with H.26L (Part 10), which, Horn notes, is completely divorced from MPEG LA's efforts.
Also, there may be questions about codecs that are billed as "MPEG-4-based" yet don't play on MPEG-4 decoders. "If they use MPEG-4, they will need our license," he says. "But otherwise you have to take a closer look at what 'MPEG-4-based' means. They might need a license for some of the underlying patents from individual patent holders."
http://www.e-insite.net/commvergemag/index.asp?layout=article&articleid=CA236131&pubdate=8/1....
culater
Join the InvestorsHub Community
Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
