# Data Compression - Systematisation

## ---- Adaptation ----

#### by T.Strutz

The **Adaptation** of the compression process plays a special role in compression schemes. Algorithms of each category (see figure 1) can be designed adaptive.
All compression techniques assume that the signal, which has to be compressed, fits a certain signal model with specific parameters. The compression works well as long as the actual signal meets these assumptions.
Otherwise the compression ratio can decrease dramatically.

Two cases can occur
- The parameters are changing during the compression. The adaptation has to observe the parameters and to modify the processing dependent on this.
- The signal model alters during the compression. The adaptation has to select an available model, which fits the signal best.

In general, the adaptation rests on observations of the past and hopes that the deduced parameters or models are also suitable (and optimal) for the processing of the present sample/symbol.
The relation between observation and estimation of parameters or models is either context-based or based on a finite-state machine.

It must be noted that it is always possible to differentiate between basic compression techniques and their adaptation. (In connection with entropy coding, adaptation is often named as "modelling".)

Strutz / 17.02.2003