A Novel Watermark Blind Attacking Algorithm Based on Spase Analysis and ICA

On the basis of the watermarking sparse feature in wavelet domain, a novel watermark blind attacking algorithm is proposed, in which the accuracy of the separation matrix was improved, by the using of sparse feature of watermarking distribution in wavelet domain, the original image is used as sparse components, the embedded watermark as noise, and sparse decomposition process associated with the ICA method. Sliding window that smooth the data is to overcome the block shadow in sparse process to enhance the intensity and scope of the attacking. The simulation experiment results show that we can on the one hand ensure imperceptible character of watermarked image, on the other hand can destroy the hiding watermarking information more greaterly, and furtherly verify the view point.