Smooth signal □
Successive differences ■
FIGURE 12.21 A smooth signal and its successive differences k
FIGURE 12.22 Sample waveform of "ae" sound as in cat
Figure 12.23 gives an example of an image where, in smooth portions of the picture, neighboring samples tend to have the same values. Long-term dependencies can also be observed when images contain periodic patterns. The compression techniques discussed in this section attempt to exploit the redundancies to attain greater compression and hence to provide more efficient representations.
Three basic compression techniques exploit the redundancies in signals. Predictive coding techniques, also called differential coding techniques, attempt to predict a sample value in terms of previous sample values. In Figure 12.21 the sequence of differences occupies a smaller dynamic range and consequently can be coded with greater accuracy by a quantizer. Predictive coding techniques are used extensively in the coding of speech signals. A second type of technique involves transforming the sequence of sample values into another domain that yields a signal that is more amenable to lossless data compression. These types of techniques are called transform coding techniques. Transform coding is used
extensively in image and video coding. Subband coding is an important special case of transform coding. A third type of technique is to design quantizers that deal not with individual samples, but instead map blocks of samples into approximation points that have been designed to represent blocks of samples. This third technique, which is called vector quantization, is usually used in combination with other techniques. In the remainder of this section we discuss the first two types of techniques.
Was this article helpful?