Chapter 4 – Basic Message Coding

Publisher Summary This chapter discusses basic methods for mapping messages into watermarks. It particularly focuses on simple methods of modifying the types of watermarking systems. The key issue in the design of a code is code separation, i.e., the distances between different message marks in marking space. If these distances are too small, there will be a high chance that one message embedded in a Work will be detected as a different message. In the case of correlation-based watermarks, the point of interest is the correlation between message marks. Message coding is often implemented in two steps: Source coding maps messages into sequences of symbols and modulation maps sequences of symbols into message marks that can be embedded in content. Multisymbol codes, in which messages are first mapped into sequences of symbols (source coding) and then systematically mapped into message marks (modulation), reduce computational complexity. Further, there are several ways a detector can determine whether or not a multisymbol message is present in a Work. These include defining a large fraction of possible messages to be “invalid” by, for instance, appending a checksum to the end of each embedded message. Any message with an incorrect checksum is invalid. If a valid message is detected, the detector declares that the watermark is present. Identifying the most likely message mark in the Work is to demodulate and decode the most likely message and then reencode and remodulate that message to obtain the most likely message mark. Once the most likely mark is identified, apply a separate test for its presence in the Work, such as computing the correlation coefficient between the mark and an extracted vector.