H.264 codec jeopardizes MPEG-4's ascendancy

H.264 codec jeopardizes MPEG-4's ascendancy
AMSTERDAM, Netherlands — Proponents of an emerging video codec called H.264 are predicting the scheme will turn the video market on its head by enabling delivery of Internet Protocol-based broadcast-quality video at data rates of less than 1 Mbit/second. Although demand for H.264 may not hit volume before 2004, the broadcasting industry's interest in the codec has gone way past the talk stage — far enough that MPEG-4, long pitched as the logical interactive enhancement to MPEG-2, could be lost in the shuffle.

The first rumblings of the promised upheaval were felt here last week at the International Broadcasting Convention (IBC). Chips, evaluation boards and software tools targeting the H.264 codec (formerly H.26L) trickled into private demonstrations and a few public venues at IBC. Presenters of H.264 previews included Canadian company VideoLocus Inc. and Germany's Heinrich-Hertz-Institut f¼r Nachrichtentechnik Berlin GmbH (HHI). Texas Instruments Inc., in launching its 600-MHz digital media processor, joined with such third-party partners as UB Video Inc. and Ingenient Technologies to show an H.264 video algorithm running on its C64x DSP family. And Equator Technologies similarly claimed its media processor's architecture will be capable of real-time H.264 encoding and decoding.

The yet-to-be-ratified international codec is now officially known as H.264 in the telecommunications world. Some in the MPEG community, however, are calling it MPEG-4 Part 10, and yet a third faction refers to it as "the proposed JVT/AVC standard."

Making waves

By any name, the codec, scheduled for completion in spring 2003, is a product of the Joint Video Team (JVT) of the International Telecommunication Union and International Standards Organization. And it's already making waves.

Many system vendors and chip companies are scrambling to devise or obtain cost-effective H.264 solutions that will work well with current system and silicon architectures, which are optimized for today's MPEG-2-centric video infrastructure.

Europeans are hatching plans to "include MPEG-4 audio and H.264 video as new options for IP-based video delivery by revising the current Digital Video Broadcast [DVB] specifications," Ken McCann, chairman of the AV-coding group at the DVB standards group, told EE Times. McCann, a director of ZetaCast, an independent technology-consulting company specializing in digital TV, said DVB's decision to go with an H.264 codec is not a done deal but "very near" to one. Unless work within JVT goes "horribly wrong," he said, DVB's AV-coding group is on its way to putting together H.264 implementation guidelines for DVB broadcast applications.

The more resources Microsoft Corp. pours into the promotion of its new Windows Media Series 9 video codec — its latest wedge to crack the professional and consumer video markets — the higher the industry expectations for H.264 as going one step beyond Windows Media.

H.264 is "no doubt the best codec there is, offering a great coding efficiency," Tim Schaaff, vice president of the interactive-media group at Apple Computer Inc., said at IBC last week.

And Nick Thexton, vice president of the TV platforms division at the News Corp.'s NDS Group, predicted that H.264 "will remove the advantages Windows Media 9 has."

H.264 proponents say the codec's ability to reduce bandwidth by 50 percent or more has the potential to permit phone companies to deliver broadcast-quality video without ripping their infrastructure apart. It could enable cable and satellite operators to offer more channels, while letting consumers store twice as many programs in personal video recorders or to record high-definition movies on DVD recorders.

Industry discomfort

Nevertheless, H.264's potential impact on MPEG-4 is causing some discomfort in the industry. The codec was intended as an extension of the MPEG-4 standard (hence the "MPEG-4 Part 10" moniker embraced by some). But with the newest video codec emerging so quickly, the budding MPEG-4 Advanced Simple Profile video spec might wind up dead on arrival or at least instantly irrelevant in certain parts of the world.

The issue is whether service providers will choose to leapfrog current-generation MPEG-4 video-coding profiles, such as the Advanced Simple Profile, in favor of H.264. Noting that H.264 is not yet a completed standard, Michelle Abraham, senior analyst at In-Stat, predicted, "The upshot is that the wait for H.264 could stall the MPEG-4 market in the broadcast space. No operator is going to switch twice."

Software tool companies like Envivio and iVast — and chip companies like Sigma Design — that gambled on current-generation MPEG-4 profiles and features may find themselves caught between a rock and a hard place. They will face pressure to come up with their own H.264 plans, well before they can start making money on new products based on current-generation MPEG-4 profiles.

Seen at IBC

VideoLocus (Waterloo, Ontario) last week demonstrated H.264 running on its own evaluation platform in conjunction with a noise reduction filtering technology developed by LSI Logic Corp. Although the two companies have no formal commercial agreements, VideoLocus' H.264 demo, presented at LSI Logic's meeting room at IBC, suggested mutual interest with respect to H.264.

Bob Saffari, senior director of marketing and applications of advanced video products for LSI Logic's broadband entertainment division, said the company's advanced audio expertise and motion-compensated temporal filtering (MCTF) technology — developed for professional video applications and deployed on LSI Logic's Domino media processor — will play "a complementary role" with VideoLocus' H.264 video codec. "We believe that the MCTF becomes even more critical for low-bit-rate applications," said Saffari.

VideoLocus' demo consisted of its own highly optimized H.264 codec, MPX, running a DVD-quality video stream at 944 kbits/s in a side-by-side comparison with an MPEG-2 video stream at 5.37 Mbits/s. VideoLocus' encoder algorithms run on a Pentium 4 platform with hardware acceleration coming from an add-in FPGA card. "By reducing the technology to hardware implementation, we believe that the credibility of H.264 has been elevated a notch," claimed Kevin Oerton, vice president for marketing and business development at the company.

HHI, a Berlin-based research institute specializing in mobile communications and image- and video-processing technologies, demonstrated H.264-compressed D1-resolution video running at 1.25 Mbits/s in software, in a side-by-side comparison with an MPEG-2 video bit stream at the same data rate and resolution. HHI's current H.264 decoder software does not run in real-time, but the institute says it plans to complete a real-time version running on a Pentium 4 platform before year's end.

Undoubtedly, H.264 coding efficiency comes at a price. Chip gate counts will likely increase at least threefold compared with MPEG-2 silicon, observers said. ZetaCast's McCann estimated that the complex H.264 algorithm requires a chip to carry out at least two or three times more decoding instructions.

As DSPs and media processors get more powerful, a software-only codec may be a logical, easy way out. But John Dunlop, senior design engineer at core developer Amphion Semiconductor (Belfast, Northern Ireland), argued that H.264's extra computational penalty is likely to require a highly optimized coprocessor for math- and data-intensive processing. "Meanwhile a RISC or DSP chip can be used for quality enhancements," he suggested. Amphion plans to launch an H.264 core in next year's third quarter.

Oerton of VideoLocus said his company's FPGA hardware acceleration "provides motion estimation, intra-estimation, mode decision statistics and video- preprocessing support. Without the hardware acceleration, a lone 2-GHz P4 requires 10 seconds to produce a single frame of H.264-compressed video," he said. "With acceleration, our encoder compresses standard-definition video at 30 frames per second."

Chris Basoglu, director of technical marketing and applications at Equator, claimed that in the architecture phase of the BSP-15 media processor, Equator put "an emphasis on high performance yet programmable motion estimation.

"We can issue 32 sum-of-absolute-difference operations every clock cycle," Basoglu said. "With the efficient use of data flow and data cache, and with programmable bit-stream parsing and generation, our processor will be able to perform real-time H.264 encoding and decoding."

'A living codec'

MPEG-4 software tool companies downplayed the possibility that H.264 will steal thunder from MPEG-4 products.

"MPEG-4 is a living codec," said Envivio CEO Jonathan Fram. "You have to stay with the learning curve." He said that Envivio's 40-odd R&D people in Rennes, France — formerly a group within France Telecom — are hard at work on next-generation codec development.

"Even when H.264 takes over, it can be folded into a rich feature set already designed within the MPEG-4 standard," said Kent Libbey, vice president of marketing at iVast. Besides audio/video coding technologies, the MPEG-4 standard covers a broad array of multimedia applications. It supports digital-rights management for protection of assets, and an MPEG-4 systems layer allows service providers and content owners to add object-based interactive features to content.

Libbey said iVast is discussing joint-development or licensing deals with small companies focused on H.264.

And both the Envivio and iVast sources said some operators in Latin America and Asia are poised to start services based on MPEG-4 Advanced Simple Profile.

See related chart