Research On DCT-Domain Digital Watermarking And Detection Scheme

Digital watermarking technology, which is the important branch of the field of information security, has been researched broadly. It is used to label the original creator's ownership or prove to the authenticity and integrality of the digital works through embedding a kind of invisibility mark into the multimedia information products, without being apperceived by people.The watermarking of digital image is paid the close attention among all digital watermarking technology. In order to applying watermarking technology very well, after analyze the basic frame, the function and technological characteristic of each component, do deep research on Discrete Cosine Transform(DCT)-domain image watermarking algorithm. Toward the digital image watermarking applied in copyright protection, the mostly requirement are imperceptibility and robustness. In order to enhance the imperceptibility, it needs to reduce the watermark embedding quantity or decrease the watermark embedding intensity. In this way, it will make the decline of robustness and watermark capacity. If enhance the robustness, the watermark embedding quantity and intensity must be increased, it will destroy the imperceptibility. Therefore, for the sake of keep balance of imperceptibility, robustness and watermark capacity, a DCT-domain digital watermarking algorithm based on Vector Quantization (VQ) compression has been presented. Experimental results show that the watermark has good imperceptibility and robustness which proved by some common image processing functions, such as JPEG compression and cropping.Besides of embedding and extraction process, watermarking detection is another key technology of watermarking system. It judges whether the dubious digital works contain watermark by certain key and algorithm. Currently, the mostly means are correlation-based detection and Gaussian distribution detection based on statistic. Contrapose the presented watermark algorithm, contrast the DCT coefficient distribution before and after watermark embedding, this algorithm affect the DCT coefficient statistical distribution. By analyzing, Generalized Gaussian Distribution fit is more exact, so non-traditional detection algorithm is more appropriate for the new presented watermark embedding algorithm.