Watermarking techniques using knowledge of host database

Abstract

Many watermarking applications deal with a database of hosts. When given a database of hosts to be watermarked, under the traditional watermarking approach, every host is independently watermarked. That is, encoding of one host does not use the knowledge of existence of other hosts in the database. If the encoder knows in advance about all the hosts in the database to be watermarked, intuitively, it has more information and hence can perform better. However it is not clear how to exploit this information and how significant is the improvement. In this thesis, we propose the notion of knowledge of hosts database and address this question: ``If the encoder has prior knowledge of the hosts database, and the detector has full or partial information of the hosts database, how to exploit this additional information to significantly enhance performance''. To handle this question, a novel approach that demonstrates the efficacy of using knowledge of hosts database during the watermarking process is proposed. The proposed approach is generic and based on this, frameworks that address the problems associated with different applications can be designed. In this dissertation three different frameworks are proposed for three different applications, namely, copy detection, retrieval systems and database watermarking. In each case, novel methods are designed to implement each framework. Systematic theoretical formulation and practical experimental evaluation is performed to validate the efficacy of the proposed frameworks.