We have seen in sections 1.4 and 1.5 that there exist two classes of source coding: lossless source coding and lossy source coding.

The aim of lossless source coding or entropy coding is to describe the digital sequence delivered by the source with the shortest sequence of symbols, usually bits. This sequence should be designed in order to guarantee the perfect reconstruction of the initial sequence by the source decoder. In lossy source coding, the aim is to minimize a fidelity criterion such as the mean square error or a subjective quality under a constraint on the binary rate.

In this chapter, we will first review in section 2.2 the different solutions to implement lossless source coding such as Huffman’s algorithm, arithmetic coding and Lempel–Ziv coding. The remainder of this chapter will focus on lossy source coding. In section 2.3, we will study the scalar quantization and the vector quantization and their respective performances. The coding techniques for memory sources such as linear prediction, scalar quantization with prediction and transform coding and subband coding will be detailed in section 2.4. Finally, we will give some application examples such as still image compression, audio compression and speech compression in section 2.5.

The run length coding (RLC) is a simple algorithm exploiting the repetition between consecutive symbols. It is efficient when the ...

Start Free Trial

No credit card required