Research On Algorithms Of JPEG Compressed Domain Watermarking And Audio Watermarking Against Synchronization Attack

As computer technology became more advanced, network developed with rapid pace, and multi-media rose on amazingly high speed, information security has come to an increasingly serious problem which can not be ignored. Therefore, digital watermarking technology turned to be available in many aspects, such as copyright protection, modification authentication, receipt checking and mark hiding.The research of watermarking technology can be parted into image watermarking, audio watermarking, and video watermarking from the carrier angle, when it can be parted into watermarking protocol, watermarking capacity, watermarking evaluation standard and watermarking algorithm from the study direction angle. This dissertation focuses on watermarking algorithm aspect, and the main work shows as following.(1) A new watermarking algorithm based on JPEG image in the compressed domain has been proposed. In the real application, digital images are always saved with compressed format. This leads most watermarking algorithms to be invalid. The JPEG image should especially be underlined and studied due to its broad application. Comply with JPEG standard, a new algorithm has been proposed, which embeds information in the quantization DCT coefficients, adopts four-code watermark with divide-modulus-get-residue embedding method to expand watermarking capacity, and uses double-embedding method to enhance the robustness of the algorithm.(2) A new audio watermarking algorithm to solve synchronization problem has been designed and implemented. In this dissertation, the biggest value is selected as the synchronization mark, which is the sum of the sequential audio data of the same numbers in the whole testing audio data. Its greatest advantage is that it can make full use of audio information itself, and it is applicable to various kinds of audio carrier. Moreover, combined with psychoacoustic auditory model, the watermark information is embedded in the frequency transform coefficient, which can influence the audio quality least. This will make the watermarked audio content to the inaudible requirement. What's more, the algorithm adopts binary image as watermark due to human vision's better error tolerance to image details.