Informed watermarking by means of orthogonal and quasi-orthogonal dirty paper coding

Abstract

Abstract—A new dirty paper coding technique that is robust against the gain attack is presented. Such a robustness is obtained by adopting a set of (orthogonal) equi-energetic codewords and a correlation-based decoder. Due to the simple structure of or-thogonal codes, we developed a simple yet powerful technique to embed the hidden message within the host signal. The proposed technique is an optimal one, in that the embedding distortion is minimized for a given robustness level, where robustness is measured through the maximum pairwise error probability in the presence of an additive Gaussian attack of given strength. The performance of the dirty coding algorithm is further improved by replacing orthogonal with quasi- orthogonal codes, namely, Gold sequences, and by concatenating them with an outer turbo code. To this aim, the inner decoder is modified to produce a soft estimate of the embedded message. Performance analysis is carried out by means of extensive simulations proving the validity of the novel watermarking scheme. Index Terms—Dirty paper coding, informed embedding, in-formed watermarking, orthogonal codes, spherical codes, turbo coding. I.