The MPEG-2 Codec - A Practical Guide to Video and Audio Compression - page 223

11.10 Profiles and Levels

The MPEG-2 standard introduces the concept of profiles and levels. A profile is a way of
describing a basic functionality or set of available tools that is easily described and unam-
biguous when mapping conformance between the encoding and decoding processes. The
levels usually relate to the size of an image raster when applied to video standards.
A combination of profiles and levels is usually laid out in a 2-dimensional table and
some cells are often left empty as the profiles and levels are not always meaningful in all
possible combinations. For example, defining a level that uses very sophisticated com-
pression tools on a very low-resolution raster is pointless.
The values in a standard would allow for video to be encoded with dozens of param-
eters that are defined independently of each other. The number of permutations would
lead to hundreds or even thousands of slightly different video-encoding formats.
A profile defines appropriate limits or bands of values. This is most obvious when
screen sizes are examined where the most common raster sizes are specified by the profile.
Sometimes the profiles are not specified by the standards organization that created
the standard in the first place. Generally, profiles and levels are determined by the indus-
trial application of the standards and are somewhat de facto in nature.
In the case of MPEG-2, some profiles and levels are defined but additional ones have
been independently specified by SMPTE. Refer to Appendix F for a detailed summary of
all the relevant profiles and levels for MPEG-2.

11.11 Scalable Coding

Scalable coding is a technique where the coded data is separated into two parts. A very
low bit-rate stream (the base signal) is created and an extra supplementary stream (the
enhancement signal) is provided to enhance the quality. For some applications, only the
low bit-rate stream is used. For example, a previewing mechanism might use a low-qual-
ity video sequence.
The high-quality data augments the lower-quality version without repeating any of
the information. It simply provides additional data so the decoder can then reconstruct a
full-resolution picture with the extra detail.
There are two approaches to this. The SNR profile creates a so-called “moderate-
quality picture” signal plus an enhancement signal with a “low-noise picture.” The spatial
and high profiles generate a low-resolution picture signal and a high-resolution picture
enhancement signal. In both cases, combining the base signal with the enhancement sig-
nal yields the necessary information to recreate a high-quality rendition or the base signal
can be used on its own.
The SNR profile is designed to allow the moderate-quality signal to be transmitted
independently at a higher power than the enhancement signal. Receivers that have a good
line of sight and therefore high-quality, noise-free reception should be able to decode the
base signal and the enhancement signal to reconstruct the best-quality signal. Receivers
that are further away on the fringe of the reception area may not be able to distinguish the