Databases Reference
In-Depth Information
14
Subband Coding
14.1 Overview
In this chapter we present the second of three approaches to compression in
which the source output is decomposed into constituent parts. Each constituent
part is encoded using one or more of the methods that have been described
previously. The approach described in this chapter, known as subband coding,
relies on separating the source output into different bands of frequencies using
digital filters. We provide a general description of the subband coding system and, for those
readers with some knowledge of Z-transforms, a more mathematical analysis of the system.
The sections containing the mathematical analysis are not essential to understanding the rest
of the chapter and are marked with a
. If you are not interested in the mathematical analysis,
you should skip these sections. This is followed by a description of a popular approach to bit
allocation. We conclude the chapter with applications to audio and image compression.
14.2 Introduction
In previous chapters we looked at a number of different compression schemes. Each of these
schemes is most efficient when the data have certain characteristics. A vector quantization
scheme is most effective if blocks of the source output show a high degree of clustering. A
differential encoding scheme is most effective when the sample-to-sample difference is small.
If the source output is truly random, it is best to use scalar quantization or lattice vector
quantization. Thus, if a source exhibits certain well-defined characteristics, we can choose
