II
Background (continued..)
Multimedia signal can, generally, be compressed due to
following factors:
-
Statistical Redundancy
-
Significant correlation exists amongst neighboring samples within a single
image or video frame; spatial correlation.
-
Significant correlation exists amongst samples of data acquired from multiple
sensors; spectral correlation.
-
Significant correlation exists amongst samples in different segments of
temporal data such as video; temporal correlation.
-
Perceptually irrelevant data exist in the signal.
-
Data with fractal nature exist in the signal; data have redundant
high-level features across space and time.
One or more of the above features can be exploited for data compression.
Figure 1 presents a system view of a codec.
Figure 1 Generic Codec System
Source Coder - compresses the data by reducing the input data rate to a
level that can be supported by the storage or transmission medium.
Channel Coder - translates the compressed data into a signal suitable
for storage or transmission.
Source and channel coders are usually separate processes, but there
are methods for combining the two processes, and tradeoff between the codecs
can lead to a simpler system; such as video conference compression standard
H.324.
Some constraints for the bit rate minimization problem include:
-
Specified levels of signal quality; usually applied at the decoder.
-
Implementation complexity; often applied at the decoder.
-
Communication or end-to-end delay.
-
etc.