Researches On Wavelet Adaptive Digital Image Watermarking Algorithms Based On Human Visual System

Security is a very serious problem in the application of network and multimedia. To solve the problem, nowadays many technologies are considered, especially for image watermarking technology, which catches a wide attention for its unique advantages in copyright protection, and becomes the hotspot of current researches. It embeds copyright information, such as figures, sequence numbers, words and images, imperceptibly into multimedia data for tracing the distribution and usage of it. The special application of digital watermarking technique requires that the embedded watermark should be not only transparent to human observers, but also robust enough to resist to different attacks.People are the leading actor of appraising image, so human vision system (HVS) theory is very important for image watermarking technology. Watermarking algorithm of good perception property can be designed by making full use of HVS.Wavelet analysis is the most widespread implement for space-frequency analysis in the world at present, and it has been widely used in image processing. The low complexity and space-frequency location peculiarity of wavelet transform are in favor of improving the performance of watermarking algorithm. Besides, as the multi-resolution analysis characteristic of wavelet transform has the compatibility with the characteristic of HVS, it is also helpful to choose appropriate embedding location and strength for the watermark.Firstly, After make an analysis for the meaning of research, current research in the worldwide, main application field and development trend of digital watermarking, this paper gave a recapitulative summarize for the basal knowledge of digital watermarking technology from its origin, basic character, sort, basal principle, common model, Assess guide line of performance, classical algorithm and so on.Secondly, this paper introduces wavelet transform definition, wavelet transform characteristics, two-dimensional wavelet decomposability of image, and the application of wavelet transform in watermarking.Thirdly, this paper introduces basal structure and characteristics of HVS, summarizes existing Just Noticeable Difference (JND) models, and classifies them by embedding object. Then, analyses the advantage and disadvantage of existing models, defines own wavelet-based luminance mask function and the texture mask function, presents their indexes and founds new JND model.Following, two novel wavelet adaptive digital image watermarking algorithms Base on new JND model are presented in this paper: The first method is an auto-adaptive image watermarking algorithm based on wavelet-coefficient. This algorithm is proposed in the paper and watermarks were embedding and extraction in the vertical sub-band after multi-resolution decomposition. Instead of using a random sequence, a visually meaningful gray image is used as watermark which is scrambled by Arnold transform. The method uses new JND model to enhance the robustness of the watermark and avoids the perceptibility of human eyes. Another method is an auto-adaptive dual-watermarking algorithm based on wavelet-block. A visually meaningful watermark and a pseudo-random sequence watermark (named Discriminating watermark and Confirmable watermark respectively in this paper) are embedded into different wavelet coefficient sub-bands through different way. In order to improve the security and imperceptibility, Arnold transform is applied to Discriminating watermark. After decomposing the original carrier image into four bands, discriminating watermark is embedded into high sub-band by using the proposed JND model to calculating every wavelet-block's JND estimate value. Confirmable watermark, which is obtained from Discriminating watermark after an especial process with pseudo random sequence, is embedded into the wavelet coefficients of low sub-band. The statistical detection principle is used to achieve the blind extraction of the discriminating watermark and the blind detection of the Confirmable watermark. In both algorithms, many parameters and the seeds of random sequence can be regarded as the secret key, anyone who does not know the secret key cannot correctly retrieve the watermark. Experimental results indicate that the performance of the proposed watermarking algorithm is invisible, robust and secure.