Research On Color Image Adaptive Digital Watermarking Algorithm Based On Quaternion Fractional Transform

With the rapid development of computer multimedia technology, digital products such as images,audios and videos can be disseminated and republished on the Internet. Because the digital products are easy to copy and disseminate, the piracy problem is becoming increasingly serious. The emergence of digital watermarks provides a solution to this problem. Some existing color image adaptive watermarking algorithms do not fully utilize color information in the adaptive process, or do not consider the holistic property of three channels of color host images during the watermark embedding. Therefore, on the one hand, this paper extracts masking properties directly from color host images. On the other hand, by using the quaternion algebra and the color image quaternion representation, the traditional discrete fractional random transform (DFRNT) is generalized for the application of color image adaptive watermarking with the consideration of the holistic property of three channels and the inherent correlation between them. The main contributions of this dissertation are as follows:(1) Based on the related quaternion theory, quaternion discrete fractional random transform (QDFRNT) is defined. Then, an efficient computational algorithm is provided by deriving the relationship between QDFRNT and DFRNT.(2) By using the human visual system (HVS), the texture, edge and tone masking properties are first extracted directly from the color host image. After that,the constraints for watermark embedding in the QDFRNT domain are analyzed to preserve the watermarking energy through the relationship between QDFRNT and DFRNT obtained in (1). Finally, an adaptive watermarking algorithm is proposed by combining with the masking properties and QDFRNT: Under the preconditions for embedding watermark in the QDFRNT domain, by adjusting the embedding strength according to masking properties, the watermark is adaptively embedded into the real parts and imaginary parts of the medium frequency coefficients in the QDFRNT domain; the watermark is blindly extracted by using the trained SVM model; in addition, the Arnold transform and the redundant embedding strategy are introduced to improve the security and the robustness. The experimental results show that theproposed algorithm is superior to the existing related algorithms in the invisibility and robustness. The fractional order and the random matrix of QDFRNT are helpful to enhance the security of the watermarking algorithm.(3) In order to obtain the adaptive watermarking strength values for each image block, the color host image blocks are first classified according to the masking properties used in 2), and then obtained their suitable generalized Richard logic curve by using the genetic algorithm to optimize the fitness function considering both of invisibility and robustness. Using the adaptive watermarking strength values, an adaptive watermarking algorithm based on quaternion discrete fractional Fourier transform (QDFrFT) is proposed. The experimental results demonstrate that the proposed algorithm has good performance in the invisibility and robustness, and outperforms some existing related algorithms.