Video Watermarking Algorithm Based On Wavelet Transform Domain Research

With the rapid development of video products both on type and quantity, it brings a lot of increasingly prominent problems, such as copyright protection, safety certification Etc. Thus, video watermarking technology is widely used in these areas. Digital video has its own characteristics, such as a large amount of data with redundancy and the imbalance distribution of moving and non-moving regions Etc. These characteristics bring many technical problems to be solved urgently to video watermarking. Compared to the static image watermarking algorithm, the video watermarking algorithm has more stringent requirements, such as higher secutity, lower complexity, higher robustness, blind extraction and so on. This thesis designs a preprocessing algorithm of watermark image according to the image encryption principle. Then using wavelet transform, singular value decomposition and human visual characteristics, it proposes two kinds of robust watermarking scheme based on the original digital video.To improve the safety and robustness of watermark, this thesis proposes a dual scrambling preprocessing algorithm of watermark image. To achieve image location scrambling, the image is divided to even blocks, and the sub-blocks are scrambled by Arnold transform at first. Then all of the pixels of each sub-block are scambled through the algorithm based on Logistic chaotic map. Finally, all of the pixels are redistributed to realize position scrambling. After that, image gray scrambling is achieved by multi-dimensional Arnold transformation and Logistic chaotic map. The results of simulation show that the scrambling effect of the algorithm is good, and the key space is large.To lower the complexity of algorithm and enhance the robustness of watermark, a video watermarking scheme based on integer lifting wavelet transform and singular value decompostion (SVD) is put forward. The video sequences are divided into corolla group at length of4, and some of these groups are selected as the embedding area according to the key. The luminance components are transformed by means of frames integer lifting wavelet with two-layer, and the sub-blocks that divided from the low frequency frame image are decomposed with SVD. Lastly, the watermark bits were embedded in the largest singular value of each sub-block by quantization methods. The algorithm realizes blind extraction. The simulation results indicate that the video watermarking algorithm not only has the property of imperceptibility, but also has good robustness against the noise attack, frame delete attack, frame replacement attack and so on.To improve the transparency and the robustness of the watermark, this thesis puts forward an adaptive video watermarking algorithm based on3D wavelet transform. Accroding to human visual characteristics, a kind of texture classification and motion detection method are designed, and a kind of improved watermarking embedding model is established based on texture parameter, motion parameter,and quantitative parameter. It is applied to the algorithm. The video sequences are divided by histogram segmentation at first. Then choose four continuous video luminance components in each scene as the watermark embedding area. The low frequency coefficients obtained by3D wavelet transform are processed by texture classification and motion detection. According to the watermarking embedding model and the results of texture classification and motion detection, the watermark bits are embedded into the digital video by adjusting the low frequency sub-band coefficients dynamically. The video algorithm can achieve blind extraction. The experimental results show that the algorithm not only has good transparency, but also can effectively resist frame shear, frame average attack Etc.A watermark image preprocessing algorithm is designed by fusing Arnold transform and Logistic chaos mapping, and it can improve the security and the robustness of watermark. The video watermarking scheme based on integer lifting wavelet transform and singular value decompostion can lower the complexity of the algorithm and improve the robustness of the watermark. The proposed video watermarking algorithm which taking advantaging of the human visual characteristics and3D wavelet transform has good transparency and strong robustness.