A blind and self-adaptive digital watermarking based on wavelet-packet decomposition

In this paper, a novel blind digital watermarking algorithm using wavelet-packet decomposition is proposed. Firstly, the original image is transformed into frequency domain with wavelet-packet decomposition by using a key that decides which direction will be further transformed, so that the sub bands where watermark is embedded are determined; Secondly, the binary watermark image with copyright information is scrambled with a pseudo-random sequence generated by another key; According to the sub bands' own characteristics, three thresholds (ire calculated,then the watermark is embedded to the selected sub bands coefficients by using different scale thresholds depending on the level decomposition. In addition, in the watermark image extraction process, the original image is trot needed. The locations where watermark information are embedded are found by using one key; and three thresholds are generated, then the watermark information are withdrawn; Finally, the watermarking image is reconstructed by using another key. The experiments' results have shown that the image embedded watermark is invisible to human eyes and very robust to various attacks.