| Digital watermarking has recently become a very active research area. Efficient watermark attacks are needed in testing watermarking algorithms. In this research, a novel watermark attack based on independent component analysis (ICA) and multi-criteria optimization is developed to improve watermarking robustness evaluation.; The new watermark attack, Semi-ICA estimation, is first developed as a nonlinear denoising algorithm to remove noises with arbitrary distribution from correlated supergaussian signals. Semi-ICA estimation comprises an orthogonal linear transformation and a Bayesian based nonlinear estimation function. The linear transformation is to maximize the dissimilarities between the original signal and noise in the sense of nongaussianity and achieve independency among components of signals. The component-wise Bayesian based nonlinear estimation is to estimate components of the clean signal in the transformed domain. Based on these two objectives, a recursive algorithm, maximizing the difference of negentropy between signal and noise and an estimation function, minimizing mean square error, are derived.; Semi-ICA estimation becomes a watermark attack if a watermark is considered as noise. The performance of Semi-ICA attack on Spread Spectrum (SS) watermarking and DCT domain watermarking are close to the denoising attacks in Checkmark package. In attacking the Image Adaptive Watermarking (IAW) algorithm, Semi-ICA attack outperforms all denoising attacks in Checkmark v1.2 package. The theorems and experiments of Semi-ICA verify that Semi-ICA is effective not only on removing noise with arbitrary distribution but also on defeating various watermarking techniques. |
