Based On Texture Analysis In Wavelet Domain Adaptive Watermark

Digital watermarking technology is one of the important methods of copyright protection. We must ensure that the watermark is invisible and the embedded watermark images have a strong robustness. Watermark invisibility and robustness is contradictory, so maintaining the balance between invisibility and robustness is the most important thing during the study.The presented three algorithms based on analysis of the existing principles of digital watermarking algorithm and the theory of wavelet-based adaptive digital watermarking algorithm. The main contents are as follows:Algorithm one is "The adaptive digital watermarking algorithm based on quasi affine transformation". First, the algorithm used quasi affine transformation to scramble the watermark image. Second, we extract a number of blocks from the original image using chaotic sequence, and form a new sub-image using those blocks, then make the sub-image go to three wavelet transformations. The encrypted watermarking image was embedded into the sub-image, we embedded watermark by modifying the singular value, then put the blocks of the sub-image into the original position in the original image. Experimental results show that this algorithm is robust and stable to common digital image processing operations.Algorithm two is "The adaptive digital watermarking algorithm based on WPT-SVD(Singular Value Decomposition) and texture analysis". the algorithm is changed based on the algorithm one. We extend from the wavelet decomposition to the wavelet packet decomposition, which greatly expanded the watermark embedding position. This algorithm also defined the texture-sensitive, embedding strength was calculated by the texture-sensitive and wavelet block energy ratio. The watermark was embedded by modifying the singular values. Experimental results show that this algorithm has high robustness.Algorithm three is "The adaptive blind digital watermarking algorithm based on texture analysis". In practice, the original image is difficult to get, so in order to enhance the practicality of watermarking algorithm, this algorithm uses the blind extraction. Embedding strength was calculated by judging the relationship among the points around the embedded point. Experimental results show that this algorithm is robust and stable to common blind digital image processing operations.The experiment shows the three algorithms in the dissertation have strong robust and some practical value.