580 Chapter 19
All compression techniques may be divided between lossless systems and lossy systems.
Lossless compression makes use of effi ciency gains in the manner in which data are
coded. All that is required to recover original data exactly is a decoder that implements
the reverse process performed by the coder. Such a system does not confuse entropy for
redundancy and hence dispense with important information. However, neither does the
lossless coder perfectly divide entropy from redundancy. A good deal of redundancy
remains and a lossless system is therefore only capable of relatively small compression
gains. Lossy compression techniques attempt a more complete distinction between entropy
and redundancy by relying on knowledge of the predictive powers of the human perceptual
systems. This explains why these systems are referred to as implementing perceptual
coding techniques. Unfortunately, not only are these systems inherently more complicated,
they are also more likely to get things wrong and produce artifacts.
19.1 Lossless Compression ...........................................................................................
Consider the following contiguous stream of luminance bytes taken from a bit-map graphic:
00101011
00101011
00101011
00101011
00101011
00101011
00101100
00101100
00101100
00101100
00101100
There must be a more effi cient way of coding this! “ Six lots of 00101011 followed by
fi ve lots of 00101100 ” springs to mind. Like this:
00000110
00101011
00000101
00101100