Research On The Key Technologies Of Digital Watermark Based On Transform Domain

The dramatic development of computers, multimedia technologies and Internet has triggered the exponential growth of digital contents that are featured by their easy duplication, storage and transmission. Consequently, copyright issues are thus accentuated and digital copyright protection has attracted attention from a variety of communities including industries, scholars, legislators and related watchdogs. It has been turned out that copyright protection of digital contents not only relies on laws and regulations but technologies. As a strong technological support for digital copyright protection, digital watermark has been a focus in the field of information security in that it is to hide signs of watermark imperceptible to human visual system into vectors of digital contents to verify ownership and perform authentication and so on.This paper, using watermark communication models as the foundation, focuses on the imperceptibility, robustness, capacity and security of meaningful gray image watermark by appling such key technologies as Gaussian mixture model, variational reference, matrix decompositions, human visual system, independent component analysis, swarm intelligence optimization and quantum secret communication to provide efficient solutions to watermarking progress and application.The researches in the paper are summarized as follows:(1)Analyze and design a watermark algorithm based on Gaussian mixture and variational referenceA watermark scheme is designed with three components that embrace Gaussian mixture model and variational reference of the parameters, watermark hiding and extracting. By applying Gaussian mixture model to description of host image and using variational reference to estimate GMM model parameters, the host image is divided into three segments via multiple description coding(MDC), among which one segment is to hide watermark and one as the reference. Then, the cover description used to embed watermark and the reference description as well as the gray image are performed contourlet transform, where the low-frequency and intermediate-frequency subbands are adopted as the embedding space for the sake of watermark robustness and imperceptibility. In this scheme, blind extracting of gray watermark is done without the participation of original host image and a(Schur decompostion) SD-based watermark algorithm is presented by considering the algebraic properties of matrixes.The experiments in this paper have demonstrated the proposed scheme can not only satisfy the fundamental demands of digital watermarks but enjoy good robustness when confronted with such frequent attacks as filters, noises, compressions, translation, rotation and cropping.(2) Study and propose an adaptive watermark scheme based on human visual systemTo construct a matrix of embedding strength, this paper suggests a self-adaptive watermark scheme that aims to analyze the response of human visual system to brightness and texture of vector images on the basis of noise visibility function. Accordingly, watermarking signals with maximum strength can be self-adaptively hidden in the embedding space according to vector images. Besides, an algorithm is designed in the light of independent component analysis to estimate demixing matrix and watermarking signals and to separate vector signals from watermarks, which has contributed to blind extraction of meaningful grey image watermarks.(3) Study and propose an intelligence watermark algorithm based on particle swarm optimization(PSO)Considering the optimization of stretch factors, an intelligence watermark algorithm based on particle swarm optimization(PSO) is proposed in which particles undertake heuristic and global searches for optimized stretch factors. Experiments have demonstrated that the proposed scheme is efficient in resisting attacks such as pepper noises, JPEG compression, histogram equalization, average filtering, scaling, shearing, rotation, translation and mirroring. This approach, faster and open to convergence, contributes to balancing watermark security and robustness.(4) Design watermark algorithms and protocols based on quantum secret communicationGiven the secure, hidden and capacity requirement of watermarks, three algorithms are designed based on quantum secret communication. The first one, based on Bell dual basis swapping entanglement, consists of four phases including initialization, interception test, watermark embedding and extracting. Analysis shows that this algorithm can resist frequent quantum attacks and enjoy high security. The second one, on the basis of quantum error-code, has two algorithms---basic stabilizer watermarking and key-controlled stabilizer watermarking. It has been proved that both perform well in embedding and extracting watermarks with their high capacities and extensibility. The third is a threshold watermarking based on quantum secret share in which Alice is the embedding part while Bob and Charlie are the extracting part. Alice divides watermarks and vectors into two, of which one is given to Bob and the other to Charlie. Neither Bob nor Charlie can independently recover watermarks but their cooperation can. Analysis demonstrates that this algorithm functions well in terms of watermark capacity and security.