Color Image Watermarking Algorithm Based On The Visual Model And Wavelet Transform

Digital watermarking technology, which is as a tool for copyright protection of multimedia datas, has been great concern. With the further development of computer network technology, digital watermarking technology in today's information society is becoming increasingly important. This dissertation briefly introduces the basic theory of digital watermarking, importantly studies the decompositions of wavelet transform and visual characteristics of HVS (human visual system),and designs different performance types of watermarking algorithms according the main research topics which are the watermark embedding and extraction.To enhance the robustness of the watermark, this dissertation proposes a color image watermarking algorithm based on image content features in wavelet domain.We use the HVS and image content features to classify the wavelet coefficients. Then on the basis that the coefficients of different categories are encoded, the introduction of intensity measure of the critical factors is to create visual threshold model. This model determines the maximum watermark embedding strength.Optimizing the extracted watermark greatly enhances the robustness of the algorithm. Experimental results show that the algorithm has strong robustness again common attacks.In order to coordinate the contradiction between robustness and invisibility and achieve blind watermark extraction, this dissertation presents a watermarking algorithm for color image guided by visual model in wavelet domain. Considering the brightness, texture, frequency, and contrast masking, a block-based visual model in wavelet domain is created. This model determined the embedding strategy and the embedding strength. Corresponding extraction strategy is simple and does not require the original image.The introduction of robust regulatory factors in the extraction algorithm adjusts the robustness of the algorithm, which allows watermarked images to adapt to different transmission channels and the network environment. Comparative experiment results show that the algorithm balances the invisibility and robustness and has stronger robustness against attacks such as cutting, superimposed noise, color processing and so on.To solve the problem that is higher modification rate and serious image distortion in color watermark image embedding technology, this dissertation presents a new embedding technology. Improved EMD (Exploiting Modification Direction) embedding algorithm can solve the problem of higher modification rate, and the advantages of full reversibility and without rounding error of integer wavelet transform can ensure the visual quality of the image. On the basis of the advantages of both, a watermarking algorithm in which a color watermark image is embedded into a color host image is presented. The embedding strategy is designed according to the visual importance, that is, watermark embed into low frequency coefficients firstly, then any remaining embed into the high-frequency coefficients, according to the different coefficients of the frequency band, the embedding formulas are different. Experiment results show that, the algorithm has good transparency, high embedding efficiency, and robust against common attacks...