The work of setting technical
standards may not sound like the most exciting research project,
but that's only if you don't think about how much standards contribute
to the smooth operation of the technologies we rely on.
Think CDs, digital photographs, the Internet. When established
and followed, standards allow whole industries to grow around them.
For those whose work contributes to major standards -- HP Labs played
a key role in developing the JPEG-LS and JPEG 2000 standards for image compression, for instance -- the payoff comes in the form of licensing
rights to their intellectual property.
HP Labs also worked on early algorithms developed by the Moving
Picture Experts Group (or MPEG 1) - which have become the standard
for moving multimedia files such as audio and video.
MPEG 21
The lab's most recent contribution to standards came from work
that researcher Debargha Mukherjee was doing independent of the
MPEG group has direct relevance for their newest standard, MPEG
21.
"I was trying to think of a universal transmission and adaptation
architecture that would work with any scalable media," says
Mukherjee, who works with the Palo Alto (Calif.)-based New Media
project team.
The challenge was to allow people to communicate and collaborate
online with devices of very different bandwidths and other characteristics.
"Existing collaboration software won't work that well if you
have one person on a high bandwidth connection and others on PDAs
and some others on laptops," he says. "You need something
else."
Better Sharing
His answer? A concept he called Structured Scalable Meta-format
(SSM), in which any scalable content transmitted between two nodes
would have a description attached to it that is derived from a fundamental
model intrinsic to all scalable bit-streams. That description would
allow the device receiving the content to download only a version
appropriately scaled by a universal adaptation engine in the network.
People on fast networks would get large, rich versions of the content
and people on PDAs would get tiny versions.
The other key was to have the transfer of the content secure. Susie
Wee in HP Labs' Streaming Media Systems group had shown how you
could stream media that was scalable and do it through a server
you didn't trust.
In Mukherjee's framework, secure transmission could be achieved
by following an extension of the same principles, but in a completely
format-independent manner. That framework envisions a secure universal
adaptation engine that would work for a wide class
of formats that exist today or would evolve in
the future, as long as the fundamental model is not
violated.
When Mukherjee presented his work to his Labs colleagues, Wee saw
its potential relevance to the new MPEG standard being created.
In July 2002, Mukherjee's colleague in HP Labs, Giordano Beretta,
presented the ideas behind SSM to an international MPEG meeting.
Although some were initially skeptical, Beretta says that members
of the standards body later "recognized the value of the fundamental
technology."
Integral Part of MPEG-21
In the end, the proposal received a lot more attention than the
researchers expected, says Mukherjee. The team was approached by
Rensselaer Polytechnic Institute about a collaboration, and then
invited to present material at two major electronic imaging conferences
this year.
The Labs team was also asked to return and present a revised version
of their ideas to the next MPEG meeting.
In the nine months since HP's initial presentations, Mukherjee's
model is likely to become an integral part of MPEG 21. That's after
Geraldine Kuo developed a software implementation demonstrating
the feasibility of a universal adaptation engine and five further
MPEG meetings attended by Mukherjee and his HP Labs colleagues Sam
Liu and Giordano Beretta.
HP Labs' Contribution
How does the SSM model contribute to the MPEG 21 standard?
Earlier versions of MPEG (there's been a version 1, 2, 4 and 7)
were more or less concerned with how to encode and compress high
bandwidth-using media like audio and video.
MPEG 21 is all about enabling transparent and augmented use of
these compressed multimedia files across a wide range of networks
and devices by using a variety of descriptors so they can be handled
in an intelligent way.
Part 7 of the MPEG 21 standard deals with Digital Item Adaptation,
which outlines protocols for the descriptions used on the content
'envelope' to enable intelligent adaptation. It is here that Mukherjee's
work turned out to be relevant.
"Think of this as putting a file in an envelope," says
Gary Sasaki, a manager in HP New Business Development who works
closely with HP Labs. "We've described how you tell a dumb
pipe that is delivering the file what it can throw away, without
it having to know what's inside," he says.
Infrastructure for Video Manipulation
As it evolves from a company known for managing still digital images
into one that manages moving images, HP has a strategic interest
in helping determine the standards for manipulating those moving
images.
And thanks in part to Debargha Mukherjee's work, the MPEG 21 standard
looks set to be useful for much more than just manipulating video.
"If you capture the right, basic information in the descriptors,"
Mukherjee says, "then you can adapt any media type to it. The
same infrastructure that you create today should work 50 years from
now."
If technical standards enable whole new industries to grow, MPEG
21 ought to be enabling industries that we haven't even dreamed
of yet for some time to come.
By Simon Firth
|
printable version
