Deciduous Le Gump was just saying what most of the MPEG committee members think about MPEG-4. In can be used in today's limited bandwidth environment, but it's not ready today. Here's what Didier Le Gall said about MPEG-4 and C-Cube...........................
sumnet.com
From the March 25, 1996 Issue of Electronic News
MPEG Plan Sparks Controversy
By Andrew MacLellan
Mountain View, Calif. - The MPEG standards committee is quietly developing a next-generation MPEG-4 standard and this week will approve a verification model whose future implications, according to some market observers, could set the MPEG-2 community on its ear.
The emergent MPEG-4 standard_the MPEG-3 specification for high-definition television (HDTV) was essentially abandoned after being folded into the MPEG-2 standard - is at least two years away from adoption but it is already stirring strong reaction, prompting one critic to call it ''a solution in search of a problem.''
Pursued in part as a low-bit rate videoconferencing specification, MPEG-4 has taken on a vastly more significant role over the last year. The shift in focus can be partially attributed to the H.324 standard developed by the International Telecommunications Union-Telecommunications (ITUT) last year. The specification, which has yet to be formally adopted, nevertheless stole some of the thunder from the MPEG-4 standard by enabling videoconferencing over plain old telephone (POT) lines with acceptable frame and bit rates.
The vacuum stirred the MPEG committee to more vigorously explore object-oriented, cross-platform delivery of real-time compressed video in a manner which would enable transmission of any type of content. The resultant verification model, one of the last steps before the standard enters a full committee review, is slated for release at a meeting this week in Florence, Italy.
''MPEG-4 is a very different kind of standard,'' said Cliff Reader, Samsung's director of strategic marketing and applications engineering and an MPEG subcommittee chairman investigating language and hybrid coding relating to MPEG-4. ''In the case of MPEG-2 and previous standards of this kind, the committee has standardized a particular type of algorithm. MPEG-4 will standardize a suite of algorithms that handle a much broader range of tasks.''
From a hardware perspective, the inherent difference between MPEG-4 and existing versions of MPEG is the complexity of the video encoder required. To borrow an analogy from Richard Doherty, director of the Envisioneering firm, ''MPEG-4 is to MPEG-2 what magnetic levitation is to the bullet train.''
Instead of coding video frame-by-frame, MPEG-4 enables coding of specific objects within each frame, adding an element of interactivity that MPEG has never before achieved. Using a C-like language, which some have loosely compared to Sun Microsystems' Java software for the Internet, MPEG-4's syntactic language cues the system hardware with an introductory handshake, essentially programming the codec to know what algorithm is being sent.
The challenge facing chip makers lies in the development of a codec that is sophisticated enough to interpret a vast array of instructions.
And with talk about incorporating both wavelet and fractal algorithms into MPEG-4, chip houses committed to the Discrete Cosine Transform (DCT) algorithm of JPEG, MPEG-1 and MPEG-2, may have to redirect their development efforts, abandoning hardwired codecs for intelligent, extendible architectures.
''Peripheral devices in the future need to be flexible,'' said Chip Stearns, S3's GM of Multimedia Products. ''MPEG-4 is very demanding in that it incorporates any kind of codec that can be downloaded. MPEG-1 and MPEG-2 standards are fixed and very well-defined, and, therefore, the hardware that supports them is fixed. This is certainly not true for MPEG-4.''
Embraced by many members of the semiconductor community, MPEG-4 does have its critics. Didier LeGall, for one, is C-Cube Microsystems' CTO and VP of R&D, and served as chairman of the MPEG committee during development of the MPEG-1 and MPEG-2 standards.
''Most of the people I know of believe MPEG-4 is a solution in search of a problem,'' said Mr. LeGall. ''In fact, it isn't much of a solution at all.''
Mr. LeGall views the MPEG-4 subcommittee's claims as fantastical, and said what actual research is taking place is being driven primarily at the behest of European governments with little or no market experience. Envisioneering's Mr. Doherty countered that Europe, with its sparse cable television infrastructure, is an ideal environment to pursue a new data transmission format.
''You don't do research in a standards committee. You do research in a research environment,'' Mr. LeGall said. ''MPEG-4, by breaking loose in a sense (from its original purpose), doesn't have a goal or a direction. They'll finalize something that doesn't make sense from an industry perspective.''
''Like Didier, I am a bit skeptical, but at the same time I think it could come up with results which could have great significance,'' added Dimitris Anastassiou, a professor of electrical engineering at Columbia University, director of the school's Image Technology for New Media center and an MPEG researcher.
''As a university, it provides a great opportunity for our students to work on these items. I think many of the results will come out in the next 10 years, not right away, but I am following it with great interest.''
Mr. LeGall agreed that the effort it takes to port a new algorithm to hardware will exceed the MPEG-4 standard adoption timetable intimated by the MPEG committee.
''The schedule is not consistent with the difficulty of the invention,'' he said. ''If you really want to do a good job at providing video compression across platforms, you have to give it time. But they are stuck with having to publish something very soon, and I don't think they have a clue.
''MPEG-4 hardware is very, very far away if it even will exist, which I doubt. But I would love to be proven wrong.''
Mr. Doherty assumes opinions such as the one espoused by Mr. LeGall stem from DCT-reliant companies who do not look forward to the proposition of redefining their algorithms.
''If 99 percent of your paycheck is written in DCT, you don't have much tolerance for this new religion,'' he said.
S3's Mr. Stearns, whose company, while not a member of the committee, supports the intent of MPEG-4, said IC makers will indeed have to redesign their codecs and perhaps face the fact that software-intensive applications such as MPEG-4 may force graphics chip houses to reinvent themselves.
''MPEG-4 is definitely being attacked from the top down, and the software is being addressed first,'' he said. ''It's up to the S3s of the world to figure out how to accelerate MPEG-4 under the software.''
''The thing is to find the right way to accelerate MPEG-4,'' added Caspar Horne, a senior member of the technical staff for multimedia house Mediamatics. ''You don't want to add a lot of hardware because that will increase cost. The key is to use the minimum hardware required.''
Mr. LeGall remains nonplused and said the MPEG-4 standard, far from arriving as a successor to MPEG-2, will only serve to confuse the multimedia industry just as it is taking off. ''All the key members of the MPEG-1 and -2 committees feel like me but in a more violent way,'' he said. ''It would be a total slap on the face to the people who spent six years time developing a standard that is very good.
''When we find a market for the problem MPEG-4 is trying to solve, C-Cube will be at the leading edge,'' he added. ''But today it's not the case.'' |
