II. Background (continued.....)

Figure 5 shows the general framework of lossy and lossless compression processing.

Figure 5 Lossless and lossy compression coding framework

Codec Classification - Implementation Viewpoint

Software-only decoders trade-off quality of the decoded image/video against the picture rate, while hardware-assisted codecs offer higher decoded image/video quality/rate at the trade-off of additional cost.
Examples of hardware-assisted implementation for CPU include Intel's MMX and SUN's UltraSparc.

Codec Classification : Sample Based or Block Based

For exploiting redundancy across time, space or frequency, codecs can use :

  1. pixel based, or
  2. block based
Pixel-based codecs operates on one pixel at a time thus can not exploit frequency information.  Block-based codecs operates on many samples at a time and can fall into one of the following categories:
  1. spatial-domain-based, or
  2. temporal-domain-based, or
  3. frequency-domain-based
High correlation in the spatial domain => High redundancy in the frequency domain, and vice versa.
For multimedia data, where the data set have different types of redundancies, hybrid of approaches can be adopted.