Digital Watermarking Scheme Based On Wavelet Analysis

With the development of society and science technology, the human society entered the informationization times. The digitized information is faced with the danger of being stolen and fabricated,etc. Specially in recent years, while the swift development of the internet expands the rights and interests of obligee, it also has brought the copyright protection crisis. Digital products can be duplicated with non-distortion, and easily and fast revised or edited by all kinds of media processing tools because the images, the videos, the audios and other works can be obtained in the digital form. These behaviors not only lay the authenticity of the works to suspicion, moreover also seriously damage the copyright owners'benefits. As an effective solution to copyright protection and information security, digital watermarking has attracted large numbers of attention in the information security field.It becomes necessary to combine digital watermarking with wavelet transform because of the wavelet favorable characteristics. Theory and experiment have proven that digital watermarking schemes based on wavelet domains have good robustness and invisibility.Wavelet theories and their application is an important subject. This thesis focuses on the construction of orthogonal wavelets and anti-symmetrical biorthogonal wavelets with finite length filters, adaptive watermarking scheme based on feature of digital image in wavelet domain and so on. The main contributions are as follows:1. A kind of orthogonal wavelets with finite length filters and parameters are obtained through the suitable selecting R-function based on the Daubechies'theory. Some examples are given.2. The construction of 4/4 and 6/6 anti-symmetrical biorthogonal wavelets with filter length (4,4) and (6,6) is studied respectively.3. Adaptive watermarking scheme based on the feature of digital image in wavelet domain is presented. The feature of the image is analyzed by using fractional box-counting dimension joining human vision system in this scheme. Scrambled watermark is two times embedded into the low-frequency in wavelet domain. This scheme can get a watermarked image with perceptual invisibility and strong robustness. The extracted watermark is the average of two extracted watermarks. The simulation result indicates that this scheme is strongly robust to different types of attacks such as JPEG compression, noise addition, cropping, filtering and so on. 4. Comparison of robustness is given when orthogonal wavelets with finite length filters and parameters and Daubechies wavelets are used in the same watermarking scheme under the condition of filter length equal. The same method is used in 4/4 and 6/6 anti-symmetrical biorthogonal wavelets. The result indicates that from generally speaking, orthogonal wavelets with finite length filters and parameters surpasses Daubechies orthogonal wavelets, and the change of parameters guarantees the security of the watermark, and optimal wavelet which is suitable to this watermarking scheme is selected. Under the same experimental condition, comprison of test results of optimal wavelet ,4/4 and 6/6 anti-symmetrical biorthogonal wavelets is given, and optimal wavelet which is suitable to this watermarking scheme under different requests is selected.