Study of digital image watermarking in curvelet domain

Abstract

The rapid development of Internet introduces a set of challenging problems regarding security. Anyone can duplicate the digital data without other permissions. Therefore, it raises the awareness of the copyright problems in the e-commerce age. Digital watermarking is a technique of embedding a hidden signal into digital contents to identify the legitimate owner, track unauthorized use or detect intentional tampering of the original data. Many research papers are proposed to embed the watermark in different frequency domains rather than in spatial domain for increasing the robustness of watermark. For the same reason, this thesis focuses on embedding the watermark into the Curvelet Transform developed by Candμes and Donoho. It is proved that one can recover the object from noisy environment by simple Curvelet wedges better than other transform. Therefore, Curvelet Transform shows great potential for watermarking. Traditionally, Curvelet watermarks are embedded into all frequency bands. However, the study of individual band or mixed bands behavior and the use of them for watermarking have not been reported. As a result, this thesis provides a detailed study of the watermarking using individual band or mixed-bands based on the Curvelet Transform. The experimental results demonstrate that the proposed algorithm can provide excellent robustness against most image processing operations including jpeg compression, histogram equalization, median filtering, Laplacian sharpening and Gaussian low-pass filtering. In order to provide better perceptibility of the watermarked image, we have added a Human Visual System model to improve the proposed algorithm. Comparisons of other watermarking methods based on the Curvelet Transform have also been studied.