An Image Watermarking Scheme Based on Information Theoretic Principles

image watermarking, information hiding, side information In this report we investigate information hiding systems with non-malicious attack channels in which the decoder does not have access to the original source data. Recent works show that such information hiding systems can be viewed as communication systems with side information at the encoder. Based on this observation, we propose a novel watermarking scheme, named scaled bin encoding, which is based on capacity achieving random encoding scheme for an additive Gaussian channel with side information. This scheme out-performs previous watermarking techniques when applied to synthetic and imagery data. Furthermore, a theoretical and computational framework for the implementation of a family of information hiding schemes is proposed. This family, named binary modulation schemes, provides a computable expression for the theoretical information embedding rate, as well as fast encoding and decoding procedures. The decoding procedure is based on the maximum likelihood principle that improves upon the standard correlation based decoding techniques. We describe experimental results of implementing various binary modulation schemes for the additive Gaussian information hiding system and for DCT domain image watermarking .