Research On Digital Watermarking Algorithm For Images Based On Wavelet Domain

With the rapid development of network and communication and extensive use of multimedia technology, the copyright protection of digital median work is becoming more and more important. Digital watermarking is viewed as an effective way of intellectual property right protection of multimedia data.Adaptive watermarking technology for image copyright protection in wavelet domain is researched in the dissertation. At first the principle, typical algorithms, attack analysis and evaluation of performance of digital watermark is described. Then, based on the research and the problems that exist to be considered and solved, two adaptive watermarking algorithms in wavelet domain for images are proposed. The two algorithms mainly focus on how to get the best embedding strength of the watermark and the scheme of wavelet coefficient modification and blind extraction. The two algorithms are as follows:The first algorithm is an adaptive blind watermarking algorithm based on Discrete Wavelet Transform (DWT) and energy model. The watermark image is scrambled by a generalized Arnold transform so as to improve the security of the watermark; the watermark is embedded into the low-frequency subband which guarantees the robustness of the algorithm; the embedding strength is calculated adaptively to improve the imperceptibility of the watermark. In order to implement the blind extraction, this paper improves an embedding algorithm based on bit-replacing, which enhances the applicability of the algorithm.The second algorithm is an adaptive blind watermarking algorithm based on wavelet packet transform and the features of human visual system (HVS). The algorithm chooses middle subband of the wavelet packet decomposition of the original image to embed the watermark, which increases the capacity of the watermark and improves the imperceptibility also. The HVS model is improved to fit for the features of wavelet packet transform based on which the watermark is embedded adaptively. The watermark is embedded by quantizing the wavelet packet coefficients, and the quantizing step is calculated according to the embedding strength.Quantity of experiment data shows that both of these two algorithms have good robustness, and the imperceptibility of watermark can be guaranteed at the same time.