Information Technology Reference
In-Depth Information
MPEG supports a variety of bitstream types for various purposes and these are shown in
Figure 1.17
. The output
of a single compressor (video or audio) is known as an elementary stream. In transmission, many elementary
streams will be combined to make a transport stream. Multiplexing requires blocks or packets of constant size. It is
advantageous if these are short so that each elementary stream in the multiplex can receive regular data. A
transport stream has a complex structure because it needs to incorporate metadata indicating which audio
elementary streams and ancillary data are associated with which video elementary stream. It is possible to have a
single program transport stream (SPTS) which carries only the elementary streams of one TV program.
Figure 1.17:
The bitstream types of MPEG-2. See text for details.
For certain purposes, such as recording a single elementary stream, the transport stream is not appropriate. The
small packets of the transport stream each require a header and this wastes storage space. In this case a program
stream can be used. A program stream is a simplified bitstream which multiplexes audio and video for a single
program together, provided they have been encoded from a common locked clock. Unlike a transport stream, the
blocks are larger and are not necessarily of fixed size.
1.13 Drawbacks of compression
By definition, compression removes redundancy from signals. Redundancy is, however, essential to making data
resistant to errors. As a result, compressed data are more sensitive to errors than uncompressed data. Thus
transmission systems using compressed data must incorporate more powerful error-correction strategies and avoid
compression techniques which are notoriously sensitive. As an example, the Digital Betacam format uses relatively
mild compression and yet requires 20 per cent redundancy whereas the D-5 format does not use compression and
only requires 17 per cent redundancy even though it has a recording density 30 per cent higher. Techniques using
tables such as the Lempel- Ziv-Welch codes are very sensitive to bit errors as an error in the transmission of a
table value results in bit errors every time that table location is accessed. This is known as error propagation.
Variable-length techniques such as the Huffman code are also sensitive to bit errors. As there is no fixed symbol
size, the only way the decoder can parse a serial bitstream into symbols is to increase the assumed wordlength a
bit at a time until a code value is recognized. The next bit must then be the first bit in the next symbol. A single bit in
error could cause the length of a code to be wrongly assessed and then all subsequent codes would also be
wrongly decoded until synchronization could be re-established. Later variable-length codes sacrifice some
compression efficiency in order to offer better resynchronization properties.
In non-real-time systems such as computers an uncorrectable error results in reference to the back-up media. In
real-time systems such as audio and video this is impossible and concealment must be used. However,
concealment relies on redundancy and compression reduces the degree of redundancy. Media such as hard disks
can be verified so that uncorrectable errors are virtually eliminated, but tape is prone to dropouts which will exceed
the burst-correcting power of the replay system from time to time. For this reason the compression factors used on
audio or video tape should be moderate.
As perceptive coders introduce noise, it will be clear that in a concatenated system the second codec could be
confused by the noise due to the first. If the codecs are identical then each may well make, or better still be
designed to make, the same decisions when they are in tandem. If the codecs are not identical the results could be
disappointing. Signal manipulation between codecs can also result in artifacts which were previously undetectable
becoming visible because the signal which was masking them is no longer present.
