Data compression Up until now we've considered systems which always tty to preserve all the information content of a message. For example, the CD digital system attempts to digitally encode information in a way which accurately represents all the nuances of l!,I!J' input audio waveforms that fits within a 20kHz passband and a 95 dB dynamic range. To do this for two channels (stereo) we record and replay 2 x 16 x 44,100 = 1,411,200 bits per second. However, as we discovered in an earlier chapter, some messages aren't very surprising (or interesting) and therefore don't contain much 'real' information. This raises two questions:

The answers to these questions are important because, if we can reduce the amount of bits required, we can send or store useful messages with equipment which has a lower capacity (i.e. cheaper!). The term Data Compression has come into use to indicate techniques which attempt to 'Stuff a quart into a pint pot'. Unfortunately, this term is used for a variety of methods which actually divide into two distinct classes. Genuine data compression methods attempt to cut down the amount of bits required without losing l!,I!J1 actual information. Other techniques, which I'll call Data Reduction or Data Thinning, seek out and discard information which they judge 'unimportant'. Data thinning does throw away some real information, but if it works well the lost information isn't missed! In this chapter we'll look at Lossless data compression. We'll consider data thinning in the next chapter.