Digital transmission technology l: source encoding - Signals, Processes, and Systems - page 340

there is information contained in audio-signals which we cannot perceive. It is irrelevant.
In this context the expression irrelevance reduction is used. We have got used to
"poor-quality" television pictures. Lossy compression processes are used here.

The differentiation between these two kinds of compression is also important. Lossy
processes compress much more efficiently than low-loss procedures. From a signalling
and theoretical point of view more lossy compression is characterised by a larger noise
element. Noise here embodies the “disinformation”, i.e. the tolerated loss of information.

A/D conversion with its part processes sampling, quantization, and encoding can be seen
as a lossy compression method (see Illustration 145, Illustration 149 and Illustration 184).
The selection of the sampling rate, the number of quantisation steps and the encoding
method have a great influence on the quality and compactness of the digitalised signal.

In order to know for what signal or type of data a specific compression process is advan-
tageous a few compression strategies will be explained using examples.

RLE encoding
RLE (run length encoding) is probably the simplest but sometimes the optimal
compression process.

Principle:

more than three identical consecutive Bytes are coded by their number.

Example :

"A-Byte" A; AAAAAA is coded as MA6. M is a marker byte and designates an
"abbreviation" of this kind. In this example there is a reduction of 50%. The
marker byte must not be present in the source text as a sign, otherwise it would
have a "double" meaning.

Application :

RLE is particularly suitable for data sets with long sequences of the same signs,
for example, black and white drawings. It is therefore frequently used for fax
formats in which large white areas are only occasionally interrupted by black
letters.

Note :

Data files with frequently changing bytes are highly unsuitable for this process.

HUFFMAN encoding

Principle:

it is based on the morse alphabet principle. The shortest codes are allocated to the
symbols which occur most often (e.g. letters) and the longest to the most
infrequent ones. The symbols of the source are coded and not the data to be
transmitted. This is called entropy-encoding. It is loss-free.

Procedure:

It must first be determined what symbols occur (for instance, in the text). Then
with what frequency (or more precisely, what probability) they occur. The