Research And Application Of Digital Watermarking

In recent years, the rapid growth of Internet and the popularization digital multimedia have made most of people use digital equipments to process, transmit and storage all kinds of information, but also have made authority and integrity of information more and more important issues. Therefore, it has become a hot research subject currently that how to resist unintentional modifications or malicious attacks well as to identify own copyright effectively. Under this background, digital watermarking as a new research area of information safety has been proposed to accomplish copyright protection or content authentication. The digital watermarking means that some digital information such as figure, serial-number, digital image, a piece of audio are embedded into the multimedia data as carrier information, not only does solve the problems such as the copyright violation of digital products, the illegal copying and tampering and the authentication of integrality, but also supplies a rational system mechanism for website protection, concealment communication and different electronic bills to against the forgery. On the base of the conventional watermarking researches, this paper has carried out further study on image and audio watermarking applying key technology, such as chaos theory, error correcting code, spread spectrum, wavelet transform. The main research works are described as follows.1. A new algorithm of embedding bilateral figures' digital watermark based on fuzzy classification and a new method of evaluating a digital image scrambling degree were proposed at one time. First, the watermark was disordered by chaotic mapping. Second, base on the Human Visual System's visual masking, an adaptive fuzzy function was used to classify the subimages of diverse texture complexity applying maximum membership principle. It can be found the invisibility and robustness of watermarking achieve the better effect so as to against ordinary image processing.2. A new multiple audio watermarking algorithm applying DS-CDMA is proposed. The main advantages of proposed algorithm include: (1) An orthogonal Gold sequence set is generated via preferred-pair of m sequences and used to spread spectrum of watermarks; (2) All the watermarks can be embedded without considering embedding order and be extracted independently; (3) DS-CDMA technology guarantees the embedding watermarks of robustness; (4) CDMA encoded watermark messages are mixed with a chaotic sequence for more security; (5) Each embedded watermark is achieved stably by using the fast fixed-point independent component analysis (FastICA) algorithm.3. Under the guidance of a general frame of visible watermark, a new visible watermark algorithm against distortions is proposed. The work was divided into two parts. First, a binary watermark identifying the copyright is visually embedded into some subblock of host image and a visible watermarked subimage is informed. Second, this gray watermarked subimage is processed by confusedly transformation and decomposing into bit planes as new watermark messages. Then they are self-adaptively embedded into other subblocks of host image for an invisible watermarking. So the visible watermark subimage can be recovered at any time in order to testify completely the copyright of digital products.4. Some robust watermarking in frequency domain were researched by using independent component analysis (ICA), back-propagation neural network (BPNN), error correction coding and orthogonal spread spectrum technology. For the proposed algorithms, simulation experiments focusing on three properties(imperceptibility, robustness, security) were done on computer. The experiment results demonstrate that the watermark embedded with those algorithms can be extracted with low bit error rate and high robustness against common signal processing manipulations.5. Two watermarking for content authentication and large-scale cropping are proposed. (1) A new audio watermarking algorithm was proposed to accomplish both copyright protection and content authentication with localization using ICA and spread spectrum technology. Compared to the conventional algorithms, this algorithm can detect the tampered areas in extracting procedure and does no longer need to embed two watermarks. (2) A classification-based scheme for image tamper and recovery is presented to resist the large-scale cropping. First, analyze the mean-values of two 2×2 blocks and classified into different kind. Second, embed the mean-value as characteristic value of each kind into the mapping kind in the image. The embedded watermark data itself is of the representative information of original image, not just of block. So it is more reliable for block recovery. The experimental results show the effectiveness and reliability of the proposed scheme. The quality of recovered image is still very high (≥30dB) when the image has been tempered 95%.