Research OnGray-level Image Watermarking With High Capacity Based OnSpread Transform

As an important embranchment in the field of information security, digital watermarking technique, which has been widely applied to copyright protection, information hiding, content authentication etc., is drawing extensive attention from academia and the industry community.Currently, most digital watermarking algorithms aim to embed binary sequences, while the schemes which are based on gray-level image watermark are only a few. However, from the aspect of capacity and recognizability, gray-scale watermarking has many advantages. The design of gray-scale watermarking algorithm is more challenging compared with binary watermarking, due to that larger watermarking information tends to degrade the host image’s visual quality and decrease robustness. Based on spread transform, this thesis proposed a high capacity watermarking algorithm, which takes gray-scale images as watermark and host image.The work of this thesis mainly includes:(1) The background and research status of digital watermarking technique are studied. A systematically introduction of the principles of digital watermarking is given. Besides, typical gray-scale watermarking algorithms are classified and analyzed.(2) A gray-level image watermarking algorithm with high capacity based on spread transform is proposed. Firstly, the gray-scale watermark image is decomposed into a series of binary digital images for the convenience of embedding. Secondly, to achieve better invisibility and robustness, Discrete Cosine Transform is performed on the host image, and only low frequent coefficients are chosen as cover vectors. Then, the cover vectors are projected to multiple orientations with orthogonal projection vectors, and each orientation is embedded with a bit plane using Dither modulation Method. Finally, reverse projection transform is performed on the watermarked vectors, which is followed by reverse DCT transform, to obtain the watermarked image. As the projection orientations are orthogonal, the mutual interference between the bit planes embedded in different spaces is prevented. At the same time, spread transform makes the embedding distortion concentrating on one coefficient spread to multiple coefficients, which helps to satisfy the peak distortion limitations.(3) Simulation experiments are carried out to demonstrate the effectiveness of the proposed algorithm, and different experiments are designed to test the influences of several factors on the watermarking scheme, including quantization step size, embedding position and DCT block size. Quantization step size stands for embedding strength, large step size decreases the transparency and enhances the robustness, and vice versa. As for the embedding position, experiment results indicate that watermarking schemes that only hide watermark information in DCT low frequent coefficients is more robust to JPEG attack. Lastly, different block sizes don’t make obvious difference to the performance of the algorithm.Research results show that the watermarking scheme proposed is with high capacity as well as good subjective quality, which can also withstand common attacks. In addition, the watermark signal can be exactly extracted without the participation of the original host image.