Blind Extraction Algorithm in Wavelet Domain Based on Structuring Observation Vectors

The blind watermark extraction algorithms based on discrete cosine transform need some the priori knowledge. To overcome this problem, in this paper, two kinds of blind extraction algorithms using wavelet transform are presented with out prior knowledge. Firstly, this paper proposed an extraction algorithm based on discrete wavelet transform. The low frequency component embedded in the watermarked image based on discrete wavelet transform, combined respectively with its horizontal and vertical components are constructed as observation vectors. Then independent component analysis is used to extract the watermark image. Secondly, this paper proposed a blind extraction algorithm based on redundant discrete wavelet transform, by watermarked image, the mixing of the redundant discrete wavelet transform component of watermarked image and random matrix, to construct the observation vector. Then the watermark image can be recovered by independent component analysis. According to experiments, the proposed algorithms have better performance and can effectively achieve blind digital watermarking extaction.