|
|
 |
|
|
[ In Focus ] |
 |
|
 |
|
 |
|
A choice of compression
|
|
by Wayne Cole
|
Now that MPEG-4 surrounds us in everything from cell phones to camcorders and surveillance DVRs, many video pros are wondering how it differs from MPEG-2 and what it can do for them.
MPEG-2 groups pixels in an image into macroblocks (typically 8x8 pixels in size) and then tracks the changes in each macroblock's position over a finite number of frames. In HDV, these groups of pictures (GOPs) consist of a reference (I-frame) and predicted (B and P) frames that are six frames (720p) or 15 frames (1080i) long.
MPEG-4, however, tracks arbitrarily shaped moving "objects" against a background plane. At its simplest, MPEG-4 considers each frame to be a video object plane, or VOP. At higher profiles, MPEG-4 describes a number of objects that exist not only over time (in other video frames) but in different planes within the frame.
At their most complex level, MPEG-4 motion vectors describe the change in each object's position, texture, and surface deformation from one VOP to another. Tracking the movement of these mathematically "synthesized" objects as a scene progresses provides better quality video at lower bit rates than square macroblock tracking.
Similar to MPEG-2, MPEG-4 uses I-VOPs (intracoded VOPs) and difference vectors to create B (bi-directional predictive) and P (unidirectional predictive) VOPs. As you might expect, I, B, and P-VOPs are organized into groups of VOPs or GOVs.
Beyond Levels and Profiles
To accommodate a variety of picture sizes and uses, MPEG-2 created decoder specifications grouped into a table of levels (rows) and profiles (columns). The level indicates resolution, while the profile indicates the complexity of decoder needed to reconstruct the imagery. HDV, for example, uses MPEG-2 Main Profile @ High 1440 (level). This indicates long-GOP video at up to 1440x1152, with 4:2:0 color sampling at up to 60 Mbps. It also happens to be the ATSC specified profile and level for 720p HD broadcasting.
Because MPEG-4 is object based, its levels and profiles define object types and object complexity. There are more than 15 object types in four classes: video objects (flat planes with changing textures), still objects (non-changing areas of texture), mesh objects (2-D or 3-D objects whose surface may deform over time), and the face and body animation or FBA objects (humanoid mesh objects that can change facial expression and posture over time).
MPEG-4 profiles describe object complexity based on whether it is a 2-D or 3-D object, transparent or opaque. There are presently 17 approved visual object profiles. Audio is treated similarly, allowing for multi-channel "moving" audio sources.
Then there is MPEG-4 Part 10, which has acquired many aliases including H.264, AVC-1, AVCHD and AVC-Intra. Part 10 focuses more on the container than the AV essence within it. It's designed to make MPEG-4 friendly to IT-centric distribution using existing network and file system architectures and protocols.
To do this, Part 10 added two "slices" (frames or VOPs) to the video coding layer (VCL). The SP (switching predictive) frame and SI (switching intra) frame provide efficient switching between bit streams coded at different data rates. FMO, or Flexible Macroblock Ordering, specifies a method of quickly interpolating a lost slice from neighboring slices at the decoder for error concealment where low latency requirements exist. But the bulk of information in the VCL will be media already encoded as MPEG-1, MPEG-2. MPEG-4, H.261, H.262, or H.263.
Container Or Content?
As with AVI and QuickTime, the words MPEG-2 and MPEG-4 can describe a container as well as content encoding. This causes confusion with videographers who think in terms of video "format" like DV or Beta SP. While AVI and QuickTime are strictly containers for transporting content (encoded with a specific codec), MPEG standards specify both containers and content encoding that can be used independently of -- or intermixed with -- each other.
For example, Sony's XDCAM EX is composed of MPEG-2 Main Profile @ High 1440 stored or transported in an MPEG-4 wrapper. This makes XDCAM EX even more "IT-centric" than XDCAM HD with its MXF wrapper.
Because MPEG-4 is exploding in the marketplace, and its video encoding has a reputation for high quality at low bit rates, marketers may exploit this container/content confusion. MPEG-4 content encoding is what provides higher picture quality at lower bit rates. The container is more like a virtual connector or a cable. Even if your editing system or desktop media player can handle the encoded content, if it isn't in a compatible container, you won't be able to "hook up" to it.
Before laying out dough for MPEG-4 or AVC devices, be sure your end-to-end process can understand the encoded content, and "connect" to whatever container, stream, or file format in which it may be stored. Otherwise, you may have great pictures you can't see or process.
Wayne M. Cole, CCV, CLVI, and member of the AGCV Board of Advisors, is also the owner of IHP, a video production company located Santa Barbara, CA. Contact him at wcole@ihpweb.com.
|
|
|
Government Video is part of the NewBay Media Community
copyright 2007© NewBay Media LLC.
810 Seventh Avenue, 27th Floor New York, NY 10019 T (212)378-0400 F (212)378-2160
|
|
|
