Research On Second Generation Watermarking Algorithm Based On Image　Feature

It is well-known that digital watermarking denotes some digital labels or signatures which are embedded into digital multimedia host data by signal processing. The major application of Digital watermarking lies in copyright protection of digital media, cecure communication on network and data hiding. In this thesis, still images are considered by using the most popular technique, and the second generation watermarking based on the feature of image and lifting-based wavelet are focused on. In order to improve the robustness and imperceptions of watermarking algorithm, the following work have been done:Firstly, a new multi-resolution image watermark algorithm is proposed which is based on wavelet multistage-lifting by analysis of the first-generation wavelet's deficiencies, such as low speed and the quality of reconstructed image depending on the way extend the boundary and so on. According to the properties of the second-generation wavelet transformation and the basic idea of traditional wavelet multi-resolution, the implementation of the proposed solution is simple and it needs only a little memory. Haar, Db2, Db4 and 9/7 were used in ours experiments. The simulation results showed that the proposed algorithm not only overcomes the deficiency of quality resulted from watermark in the traditional wavelet multi-resolution image watermark algorithm when the filter becomes larger, but also greatly improves the processing speed. The watermark contains a lot of information, moreover, the proposed algorithm has a good robustness and other advantages.Secondly, the second generation watermarking algorithm was researched based on image feature by introducing MSP corner detection algorithm and lifting wavelet. The watermark information was scrambled by using Arnold transformation, and then three-level lifting-based wavelet transformation for the carrier image was done. Corners are detected from the low frequency component of this image. Finally, watermark was embedded into the high frequency coefficients corresponding to the corners. The robustness and imperceptions of the watermark are obtained at the same time. Three-level lifting wavelet transformation for the image contained watermark was performed by using the same method in process of detection. Image corners are detected from the low frequency components. The corners are taken as root nodes for extracting all the corresponding high frequency coefficients. Inverse operation of embedding watermark was performed at those coefficients. And the watermark was obtained using the given key.Experimental results demonstrated that the proposed watermarking algorithm is better robustness than the first generation watermarking algorithm with respect to geometry attack such as RST (Rotation, Scaling and translation) and cut. .