Study Of Digital Watermarking

The chapter 1 surveys the popular technique of multimedia watermarking, introduces the basis of digital watermark, explains its background, definition, model, performance indicator, classification, existing work, including theory and algorithms and application field, such as copyright protection, digital fingerprint.In the chapter 2, the generalized Gaussian model and the Laplacian model, which are used most widely in describing the distribution of the AC DCT coefficients, are studied. After their parameters are estimated according to the maximum likelihoodprinciple for the given samples, x~2 test and KS test are done. The pre-processing ofthe samples played an important role in the tests, which included discarding some samples lying out of the range of tests and selecting appropriate division precision. The results of the tests show that the generalized Gaussian model is better than the Laplacian model in portraying the distribution of the AC DCT coefficients. An image digital watermarking algorithm in the block-wise DCT domain is presents. The algorithm is designed based on the General Gaussian Model, the special case of which can be the Gaussian Model or the Laplacian Model. The detection does not need the original images and is independence of the actual image. The result of experiment shows the algorithm to be secure against the JPEG compression attack, cropping attack and some other attacks.A self-synchronization blind audio watermarking algorithm in the time domain is presented in the chapter 3. The original audio is divided into some frames. The character of the original audio is extracted with mean filtering and is replaced by the watermark based on frame. The algorithm achieves self-synchronization by searching all sampling points of the current frame. The detection does not need the original audio. The idea of the algorithm is general and can be applied into other field, such as image watermarking.In the chapter 4, the communication model of digital watermarking is investigated from the requirement of the imperceptivity in digital watermarking. The concepts of image space, perception space and watermark space are proposed and the image threshold, perception threshold and watermark threshold are defined on the three spaces. The relationship among the sender, the attacker and the receiver is analyzed among the three image sets derived from the three thresholds. An information-theoretic model of digital watermarking is presented. The model is analyzed in detail and illustrated by an example. The results show that the key ofdigital watermarking is to find a covert channel, which is similar to many information-hidings, such as steganography. Some suggestions for the sender, the attacker and the receiver are given. In this chapter, we also pay attention to steganalysis and image quality assessment, as somewhat of the attack and defend in digital watermarking.In the chapter 5, the mechanism of stochastic resonance is analyzed in detail. The applications of stochastic resonance in the analog signal and digital signal processing are investigated and an algorithm of digital image watermarking is designed based on the stochastic resonance. The watermark is viewed as the weak signal and the DCT ac coefficients of the image are viewed as the noise. The watermark, the coefficients and the bi-stable nonlinear system make up a stochastic resonance signal processor to detect the watermark. Numerical simulations results show that the algorithm can resist against some attacks, such as JPEG compression, adding Gaussian noise, histogram equalization and so on in the condition of keeping the watermark imperceptible.Some discussions and conclusions, as well as some open issues and possible future research directions, are given in the chapter 6,...