The Modulus ME1000 SD MPEG-4 AVC encoder delivers video at half the bit rate of MPEG-2.
MPEG-2 video compression has been around for about 10 years, but a compelling new alternative known as MPEG-4 AVC is now available. The advantage of this standard over MPEG-2 is increased efficiency. It delivers video at half the bit rate of MPEG-2.
The standard was jointly developed by the ITU-T and ISO/MPEG standard committees and was ratified in May 1993. Based on MPEG-2 foundations, it is designed to carry signals over the existing MPEG-2 transport and modulation infrastructure. The encoding and decoding elements of the standard are complex and technically challenging, ensuring that it will improve over a substantial period of time before the technology matures.
There are many profiles and levels associated with the standard. It is also known as MPEG-4 part 10 and is quite different from MPEG-4 part 2, a technology that failed to deliver adequate gains over MPEG-2. The standard also is sometimes referred to as H.264, which refers to the ITU-T H.264 video compression standard.
Both AVC and MPEG-2 rely on the principle that occasional pictures compressed spatially are interleaved with pictures that predict and describe the motion.
Enhanced prediction modes
Products based on MPEG-4 AVC are now ready for deployment, enabling operators to deliver SD video at significantly lower bit rates. The following is an outline of how the standard evolved from familiar MPEG-2 methods and describes the practical technology required to deliver both SD and HD AVC content.
The superiority of the standard over MPEG-2 is largely due to substantial improvement of the motion compensated prediction. First, it doubles the accuracy of the motion prediction by implementing quarter pixel interpolation. The scenes that most challenge MPEG-2 are handled more effectively with AVC. Fixed block sizing for motion-prediction processing is fundamentally a compromise. MPEG-2 relied on motion prediction based on fixed 16×16 blocks. The standard improves on MPEG-2 by supporting an adaptive hierarchical scheme with block size options down to 4×4.
Shown here is the architecture of the Modulus ME6000 HD MPEG-4 AVC encoder. The unit supports advanced features such as CABAC entropy coding, macro block adaptive field frame coding and a de-blocking filter. Click here to see an enlarged diagram.
Improved intra-prediction modes
The standard also extends the adaptive field or frame encoding mechanisms. MPEG-2 uses picture adaptive field or frame coding. AVC adds the tools to allow the field or frame coding to be adapted on a macro-block basis.
The standard also supports new intra-prediction modes that allow the spatially predicted frames and intra-coded macro blocks to be described with fewer bits, improving compression and providing more consistent processing, especially in flat areas.
AVC deserves the attention it has been getting, but it is not immune to the usual hype that surrounds a new technology. In the near term, it is realistic to expect the standard to match high-quality MPEG-2 at half the bit rate. However, AVC is poised at the start of its technology lifecycle. Its tools offer an array of permutations and development possibilities for improving video compression that are only beginning to be explored and exploited.
The AVC algorithms are flexible, with tremendous potential to be refined and enhanced. This favors a programmable approach to encoder design. Even though processing support for transform and rate control are well served by general purpose CPUs, the block-based processing and search mechanisms are not. DSP chips can be used, but a more powerful and flexible approach is to employ programmable hardware such as an array of Field Programmable Gate Arrays (FPGAs). Implementations that combine FPGAs and general purpose CPUs offer the ideal combination of processing efficiency and power to fully support AVC encoding.
Enhanced entropy encoding
The standard accomplishes this with enhanced versions of proven MPEG-2 mechanisms. The AVC standard does not define the process for managing quantization and optimizing the efficient use of available bits. General- purpose CPUs also can provide the processing cycles to support the de-blocking filter function.
Entropy coding is the final stage in the compression process. AVC Main Profile introduces powerful new technology options, known as Context Adaptive Variable Length Coding (CAVLC) or Context Adaptive Binary Arithmetic Coding (CABAC). The CABAC option is complex to implement but yields remarkable efficiency capability.
One alternative from the PC world that has survived and has received plenty of attention is Microsoft Windows Media technology. Microsoft has heavily promoted its Windows Media technology as the new compression format for video across all devices, not just PCs.
Extending AVC's reach with new profiles
Microsoft is in the process of standardizing this format, now known as VC1 through SMPTE. As a result of the standards effort, Microsoft has revealed that the underlying technology is quite similar to AVC. VC1 differentiates itself by adopting a toolset that is optimized towards decoding on PCs at the cost of omitting some tools such as CABAC that would enable greater bandwidth efficiency.
AVC is highly versatile with many profiles and levels to suit specific applications. Main Profile at Level three (MP@L3) is the appropriate one for the delivery of full resolution 4:2:0 interlaced video, while Level four (MP@L4) is designed for HD. At a meeting in Seattle last July, the MPEG committee approved a number of important High Profile amendments, known as Fidelity Rate Extensions (FRExt). These extensions enhance efficiency and the scope of the standard by offering higher fidelity options, such as 4:2:2 coding, and greater sampling range to further extend the scope and value of AVC.
MPEG-4 AVC has shown significant efficiency gains over MPEG-2, an advantage that is compelling for operations where bandwidth is limited or expensive. AVC effectively extends many MPEG-2 principles but is a big evolutionary step.
Early implementations may cost incrementally more than legacy MPEG-2 solutions, but open standards, market competition and the emergence of powerful integrated silicon will quickly take AVC beyond the technology limits constraining MPEG-2. Encoder and professional decoder products have been launched to deliver the standard over existing MPEG-2 transport infrastructures.
After many years of false starts, HDTV is now overcoming many of the obstacles that have hampered its introduction. HD content is now abundant, and HD-capable receivers are attaining mass market price points. Most importantly, the major cable and DTH operators in the huge U.S. market are finally promoting HDTV.
AVC platform implementation
HD is also strongly positioned in Australia, Korea and Japan, and strong European interest resurfaced again at the IBC2004 conference in Amsterdam. Interest in AVC is timely, as it dramatically enhances the practical ability of operators to deliver HD and other bandwidth-intensive services profitably.
Practical AVC encoding and decoding solutions must overcome substantial technical challenges. SD AVC compression requires around 10 times more processing power than MPEG-2, and there is now a dearth of ready-made silicon solutions for professional AVC compression applications. HD encoding compounds the challenge.
Despite the challenges, AVC products are being delivered. Look for SD products that can support full resolution processing and all the tools that make the standard efficient; including CABAC, macro block adaptive frame/field (MBAFF), de-blocking (loop) filter and multiple reference frames.
Neil Brydon is the director of product marketing for Modulus Video.