End of the Ride?

 Last update: 2005/03/08

The 15th anniversary of the first MPEG meeting in Ottawa, ON is an excellent time to ask the question: does MPEG still have a role to play? This is a reasonable question and I can see three reasons why it may be a good idea to (plan to) end this 15-year long ride on the media bits. The first is that, after a certain period of time, organisations tend to become sclerotic, bureaucratic, procedure-driven. In other words they tend to forget their raison d'être by confusing the means to reach the goal with the goal itself, they become incapable of reinventing themselves and do not look at the world with the same curiosity and drive they had when they were first established. The second could be that much has changed in the last 15 years and the recipes that were good 15, 10 and maybe even 5 years ago do not necessarily keep their validity today. Lastly there may not be enough standards work left to justify the existence of the group.

I do not think that a  new body with a mandate similar to MPEG's would be an advantage. MPEG has a recognised brand, a considerable experience and an established work method. Dissolving MPEG and creating a new group would mean that many industry users would lose a reference while the new group would have to fight for recognition in a sea of competing groups, all marketing their results as the solution for the needs of the DM industry at large, when they are in fact solutions targeted to a specific industry.

I also do not think MPEG is a sclerotic organisation. Let's first consider the membership. During the last 15 years at every new project, waves of new participants have come one after another to replace portions of previous generations of participants. Today less than 5% of regular meeting participants have been in MPEG for more than 10 years and the oldest participant is, ahem, the Convenor, followed by Ajay Luthra of Motorola whose first meeting was Livingston, NJ in February 1989. All chairs have changed more than once. Maybe a quarter of current participants have been in MPEG for less than a year. 

Let's now consider how vital the group has been in reinventing its role. Starting with MPEG-1 and MPEG-2 from a traditional approach to solve the needs of rather traditional users, with MPEG-4 MPEG has been able, certainly not to "forecast" the future, but at least to create future-proof technologies that are gradually being adopted as open solutions for communication over the unstructured world of the fixed and mobile Internet. Remarkably, MPEG serves new constituencies, without stopping to cater for the needs of its older constituencies. With MPEG-7, it has anticipated the needs of the next phase of media experience and with MPEG-21 it is providing the technologies that can conjugate civil and economic rights. Sclerotic? Dunno, maybe the Convenor is. Certainly not MPEG.

Let's now consider the worst that can happen to human organisations, i.e. to stop being driven by a sense of mission and start being driven by procedures instead. This sentence should not be taken to mean disdain for the rules. A body like MPEG, producing standards with such impacts on manifold industries and end users cannot afford to be loose in the way it performs its function. But it is one thing to concentrate on procedure and forget about the content of the work and another to be driven by the work and make sure that procedures are upheld to high standards to the satisfaction of all parties involved.

Hoping that the first test has been passed, let's consider the second, i.e. whether MPEG is still a match to the current challenges. It is true that the last 15 years have brought an incredible amount of changes. Audio and video used to be the realm of a restricted number of industries, while now almost any industry claims to have a stake in Digital Media. Standards used to be bound to rigid hardware implementations while now standards have to deal with the fact that their implementations are by and large pieces of computer code. Indeed, to get to the point it is now, MPEG has had to change its skin several times. From MPEG-1 and MPEG-2 where software was a development tool, to MPEG-4, where software was elevated to the rank of tool to express the standard itself, to MPEG-7 which is definitely an IT standard although developed to serve the needs of the AV industry, to MPEG-21 which is more a framework than a traditional standard, MPEG has been able to adapt itself to continuously changing environments. A side issue as it may be considered, MPEG has changed its working methods much before any other group to augment its productivity and was probably the first group of its size to make massive use of advanced ICT in standard development.

This should not be taken to mean that MPEG has already adapted itself so much that it can now take some rest. Far from it. Challenges abound and discussions are being made on how to cope with the most important of them:

1. MPEG has evolved to address a wide range of technologies. Each of these require a high degree of in-depth expertise, but the different technologies are not independent. How can MPEG retain its ability to develop standards that are individually high-level and still fit perfectly in a complex picture? One response to this question has been tried with the creation of the Integration Subgroup less than a year ago. The other subgroups will continue to provide in-depth expertise, while the new subgroup will provide a more "diffuse" expertise by integrating the reference software and conformance of the different parts.
2. MPEG has traditionally produced standards rich in functionalities but with a rather high entry level. This was the right thing to do in times when the media market was "rich". Today, however, the DM landscape does not offer much fat and in many domains users tend to be content with quick and dirty solutions. How can MPEG recover these customers without abdicating its traditional role? One answer is the creation of entry-level use cases for each standard, an example of which is the definition of an MP4 File Format that incorporates ID3 metadata
3. The practical exploitation of the MPEG standards is becoming difficult because of the conflict between the manifold ways users of the standards conceptualise their market exploitation models and the rather conservative approach that rights holders, some of them from "old" industries, take vis-à-vis the exploitation of their rights. As a matter of fact this is not something that MPEG is entitled to deal with, but it is clear that, unless some new ideas materialise, it is going to be more and more difficult for users to exploit MPEG standards (and in general open standards) and for rights holders to be remunerated for their IP.

The issue of how much need there is for a body like MPEG to cater to its digital media constituencies is one that has occupied me for some months. I have made an inventory of which technologies MPEG standards offer to users and discussed the results with the Chairs and the HoDs. The result can be represented by the following table, where regular font means technologies that are already available from published standards and italic font means technologies under development.

From the table one can see that MPEG has already produced a large number of standards with a wide coverage of technologies. Therefore there could be some grounds in an argument pointing to at least a diminished role for MPEG. In the following I will provide a list of issues that, although partial and without a quantitative assessment of the amount of work required, should give an indication of the relevance of the issues that remain to be tackled. 

The second area of endeavour is to complete the MPEG-4 and MPEG-7 projects. For MPEG-4 we are at the last few amendments, while for MPEG-7 we are still developing several parts of version 2.

The third area of endeavour is to bring to conclusion the MPEG-21 project. So far 3 parts have been completed, but as many as 9 parts are still under development. More parts, like e.g. ER are still to be started.

The fourth area of endeavour is exemplified by the recent cases of successfully completed new standards like MPEG-4 AVC and AAC Bandwidth Extension (BE). I would like to add some words to highlight what they mean for the future of MPEG.

When MPEG started its MPEG-4 project, ITU-T was repeatedly invited to develop the new standard under a similar collaborative arrangement as MPEG-2 Systems and Video. ITU-T did not respond to the offer and the two bodies went ahead independently with ITU-T adding extensions to H. 263 in the form of Annexes aimed at improving coding performance of the basic standard and MPEG building MPEG-4 Video starting from H.263. 

The two projects obviously diverged but it was thanks to Gary Sullivan, then of PictureTel, Chair of the Video Coding Experts Group (VCEG) of ITU-T SG 15 and regular MPEG attendee, that at least some elements of interoperability could be recuperated. The result is that today MPEG-4 Video is natively capable of decoding a basic form of H.263.

In the mean time VCEG went to research mode to develop a new coding technology from scratch and, towards the end of 2000, started showing interesting results. At the Pisa meeting in January 2001 I asked Gary to take the place of Thomas Sikora as Video Chair, because Thomas had changed jobs and he was appointed at the following Singapore meeting in March 2001. At that meeting a CfE was issued to acquire evidence of sufficient progress in video coding that would justify the development of a new standard. At the following Sydney meeting in July the submission made by VCEG showed good results and was taken as a basis for a joint ISO-ITU project. This time the project did not target "Common Text" as in the MPEG-2 case, because keeping synchronisation of the work in the two bodies had proved a nightmare, but "Technically Aligned Text". This meant that the standard produced by the joint project would materialise in two separate documents - one in ITU-T and one in ISO - but the two documents would be technically equivalent, even though not necessarily identical.

The collaboration framework was agreed at the Decembre 2001 meeting in Pattaya with the establishment of the Joint Video Team (JVT) chaired by Gary Sullivan. The joint work advanced in great strides and the standard - MPEG-4 part 10 Advanced Video Coding (AVC) - has been approved at the March 2003 meeting in Pattaya. Because of the workload, in March 2002 Jens-Rainer Ohm of University of Aachen took over from Gary as MPEG Video Chairman. 

A similar process, although all within MPEG, happened for Audio. A CfE was issued at the La Baule meeting in October 2000 and a CfP in March 2001. Amendment 1 to MPEG-4 Audio "Bandwidth Extension", approved in March 2003  provides improved quality compared to MPEG-4 AAC at the same bitrate, with the advantage that existing MPEG-4 AAC decoders can still decode the new BE bitstream. A similar BE work is being conducted with parametric audio coding.

The same is happening again with a new project on Scalable Video Coding (SVC) that has been started with the issue of a CfP. Indeed ten years after MPEG first started looking for the scalable video Eldorado, there is now sufficient evidence that in these 10 years research has not passed in vain. Since the July 2001 meeting MPEG has been carrying out an SVC exploratory activity with a CfE issued in March 2003. The results analysed at the July 2003 meeting in Trondheim brought evidence that a quality comparable with the one obtained with AVC can be achieved with SVC algorithms. 

But, be relaxed: there is no danger that users will be flooded with new video coding standards in rapid succession and will be left wondering which one to choose. The first reason is because an SVC algorithm is required only for certain classes of application while many others are well served by the existing set of MPEG standards: MPEG-1, MPEG-2, MPEG-4 SP and ASP, and AVC. The second reason is that development of a new standard, probably based on a radically different new technology is not going to be a simple thing. MPEG prides itself on the production of standards that are efficient in compression and practically implementable. Very likely there will be a completely new field of technologies providing scalability for which the usual compression efficiency - implementability trade off is to be discovered. A minimum estimate of the time it will take to develop the new standard is 24 months, but probably more will be required.

This scalable video and, possibly, its audio equivalent, would be a very significant complement to MPEG-21, whose paradigm is "content adapted to users, terminals and networks and scalability is certainly a feature matching the last two.

To the same category belongs the AFX standard (part of MPEG-4), recently approved after 3 years of development. AFX is the first component of a suite of technologies that has the potential to introduce a new dimension in the creation and fruition of 3D worlds.

A sixth area of endeavour is offered by the consideration that, in its 15 years of history, MPEG has contributed many means to provide richer media experiences using digital technologies. However, it should be clear that we are still far from reaching the point where the richness of AV information generated either by the real world or by synthetic means will match the processing, understanding and enjoying capabilities of the human brain. MPEG has probably something to say and to do to advance mankind toward that point. The first step in this ambitious project can be to take the results of the investigations carried out during the last 15 months on 3D Audio and Video (3DAV), into an NP that could be called "Coded representation of 3D Audio and Video information". Industrial interest in this project is building and recent news seem to indicate that the time might be ripe for a standard in this area.

Copyright © 2003 chiariglione.org