HenningerDC - Advances in MPEG-4/H.264AVC Encoding... The Mustang ASIC - February 2008
In February, about 20 SMPTE-DC members and guests gathered at Henninger Media Services to be edified by Thomson senior account executive and SMPTE-DC Manager Rudy Niznansky (SMPTE-DC manager on the benefits of MPEG-4 AVC and the enhancements Thomson is bringing to its implementation.
Uncompressed motion images require huge files and need distribution bit rates much too high for acceptably priced consumer equipment and media. Therefore, it is necessary to compress files and reduce bit rates to get the programs to homes.
Enter the Moving Picture Experts Group, an ISO/IEC working group that develops open international families of video and audio standards, including those for codecs, for the delivery of multi-media. MPEG continues to have a profound impact on moving-picture content distribution via satellite, over-the-air (OTA) terrestrial broadcast, physical media, and cable/optical-fiber/Internet streaming. There are currently five categories: MPEG-1 — the Video CD format; MPEG-2 — adopted for OTA broadcast, cable and DVD; (there is no MPEG-3); MPEG-4 — a more efficient algorithm for multi-media applications; MPEG-7 — focusing on multi-media content metadata; and MPEG-21 — concentrating on interactivity.
When DVD and OTA DTV standards and procedures were being developed, the most advanced video codec that was adequately developed, standardized and implemented was MPEG-2, leading to its adoption.
The rapid growth in demand from suppliers and consumers for SD and HD programs to be wired/wirelessly downloaded via the Internet to homes and mobile devices, and transferred across home networks, has challenged the bandwidth of distribution paths, driving the need for more efficient video compression codecs. As Niznansky pointed out, “The formula for the digital age: Bits=$$. No bits= No Bucks.”
This has led to the MPEG-4 family of standards, designed for multi-media and IP streaming applications. It is generally acknowledged to be superior to MPEG-2. MPEG-4 compression yields file sizes and bit rates less than half those resulting from MPEG-2-encoding of the same sources for a given image quality; from another perspective, MPEG-4 delivers much higher image quality than MPEG-2 for the same file size and bit rate. With other parameters held constant, MPEG-4 “delivers content faster, cheaper, and/or of higher quality” than MPEG-2. It is not atypical for a program encoded in MPEG 4 to offer the same, or sometimes higher, image quality at less than half the bitrate of the same program encoded in MPEG-2. Direct-broadcast satellite providers have been migrating from MPEG-2 to MPEG-4 over the past several years. The movie mode on some consumer camcorders uses MPEG-4. MPEG-4 is one of the codecs supported by the Blu-ray and the HD DVD high-definition disc formats. Everyone wants more channels of the same video quality in the same bandwidth.
Niznansky explained that the MPEG-4 family tool kit includes standards for AVC (ITU H.264) video compression, 2D/3D graphics, Advanced Audio Codec (AAC), Java script interactivity, XMT authoring, and the *.mp4 transport.
MPEG-4 is object-based compression (the image is broken down into various parts, or objects), which offer higher compression efficiency to objects that allow it. The composite image is created in the receiver, which puts the objects back together. This is in contrast to the MPEG-2 image-frame-based concept, in which the entire image is compressed as a single entity. MPEG-4’s object-oriented nature facilitates easier switching between broadcast and IP distribution of compressed content.
All the MPEG video algorithms result in lossy compression — data that (hopefully) is not visible is thrown away, never to be recovered. MPEG algorithms are also asymmetric —the encoder is much more complex than the decoder; encoders (of which there will be relatively few) are expensive, decoders (in every playback device) are relatively inexpensive. In addition, the standards are playback standards, implemented in the decoder. Encoders can incorporate any algorithm, any philosophy, and any technology, as long as the resultant file/bitstream can be properly decoded by an MPEG-compliant decoder. “Speed, quality, tricks and cost are up to the encoder manufacturer.”
MPEG video algorithms are scalable — any MPEG algorithm can deliver various levels of image quality depending on how aggressively it’s applied. Any program in 1080i30 or 720p60 can be called HD, even if the number of pixels per row is reduced or the image quality is execrable; it’s only the number of rows of pixels that defines HD, according to the Consumer Electronics Association.
The viewer will see less objectionable image degradation as the decoder is bit-starved. With MPEG-2, a bit-starved decoder will quickly be revealed in macroblocking, while with MPEG-4 a decoder under the same condition will first deliver a softened picture, and only when the bit rate is reduced much further will blocking occur.
Encoder implementations vary widely. For example, Thomson’s professional-grade ASIC-based (application-specific integrated circuit) encoders yield about 30% lower bit rates for the same source program as their consumer-grade ASIC-based encoders. This is because some of the MPEG-4 AVC encoding tools that Thomson has developed are reserved for professional encoders, which cost more — reasonable for a device that delivers higher performance. It requires Thomson’s entire AVC tool set to deliver the maximum MPEG-4 rate advantage over MPEG-2. Implementations also affect latency, a factor in near-real-time processing.
MPEG-4’s efficiency also means that with less aggressive compression, while still yielding a lower bit rate than MPEG-2, can more easily retain satisfactory image quality while being concatenated through multiple encode-decode cycles, and transcoded from MPEG-4 encoding to MPEG-2 encoding. It helps that since MPEG-4 has MPEG-2 in its heritage, the MPEG-2 tool set is a subset of the MPEG-4 tool set, making transcoding easier.
SMPTE VC-1, based on Microsoft’s Windows Media video codec, is a competitive video compression algorithm standard that is considered approximately equal in efficiency to MPEG-4, with performance differences dependent on image content and the degree of data rate reduction desired. VC-1 and MPEG-4 are each subject to their separate royalty and licensing requirements.
For further details, refer to the MPEG Industry Forum’s 2005 white paper: “Understanding MPEG-4: Technologies, Advantages, and Markets” (www.MPEGIF.org). Note thatwww.MPEG.orgis not an official MPEG organization site.
SMPTE-DC wishes to thank Henninger's Rob Henninger (CEO & President), Van Bond (Director of Sales) and Dae Komes (Chief Engineer) for their hospitality.