An Information-Theoretic Model for Dispersed Ordered Dithered Halftone Image Watermarking by Hilbert Curve Decomposition

The decoding error rate of watermarking codes on the binary symmetric channel with bit-error probability f is an important performance index of watermarking system. The current information-theoretic models are mainly for gray scale images. This paper provides a theoretical model to analyze the watermarking code for dispersed ordered dithered halftone images using Hilbert curve decomposition. The analytical model regards the watermarking as a binary coding process in which the sub images, derived by the Hilbert curve decomposition, are represented as samples of two independent binomial random variables. Then it uses Binomial to derive a closed expression of the detection error rate in terms of f and capacity. It is shown that the analysis and experimental results agree well.