Image Processing Reference
Version 2 of the MPEG-4 standard introduces shape-adaptive DCT coding. This is used
when computing alpha channel-coded boundary macroblocks that may only partially
cover the area occupied by the macroblock. By adapting the shape of the DCT window to
better fit the area of interest, the amount of source data to be considered will be reduced,
which improves the compression efficiency.
Profiles and Levels
The same philosophy regarding profiles and levels described earlier for MPEG-2 has been
adopted for MPEG-4.
There are two profiles that have become focal points for compression for Web and
mobile content, the simple profile (SP) and the advanced simple profile (ASP). The simple
profile is intended for low bit-rate situations and the advanced simple profile has more
advanced tools that improve the quality, but it isn't as well supported. Refer to Appendix
G for more information on MPEG-4 profiles and levels.
Professor Jens-Rainer Ohm from the University of Aachen in Germany has devel-
oped a useful taxonomy. Other useful material on the subject has been developed by Ben
Waggoner and published in DV magazine. See the appendices for a bibliography.
Professor Ohm's taxonomy divides the profiles into six regions on two axes. In one
axis the profiles deal with rectangular or arbitrary shapes. The other axis is concerned with
scalability. Here are the six regions:
Arbitrarily shaped with temporal scalability
Arbitrarily shaped with but with no scalability at all
Arbitrarily shaped with spatial and temporal scalability
Rectangular with temporal scalability
Rectangular with no scalability at all
Rectangular with spatial and temporal scalability
The standard is still evolving. There are profiles for studio systems and additional
profiles optimized for 2D support. The natural visual profiles are used for conveying
film or video content that has been digitized and will be viewed in a somewhat
Refer to Appendix G for more detail about the range of parts, profiles, and levels for
MPEG-4 part 2 and MPEG-4 part 10, also known as H.264.
At the time when this codec was being launched, the licensing terms were not attractive
to broadcasters and therefore there was little incentive to take this up as an alternative to