The example systems described in the previous chapter embed only one bit of information. In practice, most applications require more information than this. Thus, we now consider a variety of methods for representing messages with watermarks. In this chapter, we concentrate on methods that result in simple extensions of the watermarking systems described so far and that are mostly straightforward applications of methods used in traditional communications. The next chapter focuses on optimal use of side information (see Section 3.3.2), which entails a less conventional coding method.

Although we do not intend this chapter to serve as a comprehensive introduction to communications theory, we have written it assuming the reader has no prior knowledge of this field. Those who are familiar with the field may wish to skim.

We begin, in Section 4.1, by examining the problem of mapping messages into message marks (i.e., vectors in marking space). The simplest way to do this, which we refer to as direct message coding, is to predefine a separate reference mark for each message. However, this approach is not feasible for more than a small data payload. Thus, we discuss the idea of first representing messages with sequences of symbols, and subsequently mapping the symbol sequences into vectors in marking space. The section ends with an example of a simple watermarking system that uses blind embedding to embed 8-bit messages.

The performance of the system that ends Section 4.1 is not as good as we might want, and...