A Geometrically Robust Digital Image Watermarking Based On Harmonic-Fourier Moments

With the development of computer network, especially with the popularization of Internet in the world, computer network has been used in various fields in our daily life. But the accompanied problem is also serious: software resource and digital information can be easily copied, tampered and damaged illegally. On this background, it is an important task to protect the safety of multimedia digital information in the digital age. Subsequently, digital watermark technology comes into being and develops rapidly.Digital watermark is an important means of protecting the copyright of digital media in cybertimes. The digital watermark is to embed some information which can prove the copyright ownership or track tortuous message in the protected digital object (e.g. image, video). The information may be the author's serial-number, significative image or text. The watermark will not influence the effect of the original data, and it can be partially or completely recovered from the mixed data to protect the copyright.A perfect digital watermarking algorithm should have several general qualities: intangibity, rubustness, security and testability, etc. At the same time, it should have dependable detection mechanism and acceptable computing cost. A digital image watermarking system has three main parts, i.e. generating part, embedding part and extracting part. In conventional practice, the generating part and the embedding part determine the key in extracting part, and the work of embedding and extracting is symmetrically reversible (e.g. DCT transformation and its inverse transformation).Digital watermark has a wide application, and image watermark is an important part of watermark field. Geometric attacks often desynchronize the watermark detection that many watermarking systems fail. Rotation, scaling, adding noise and JPEG compression are common geometric attacks. Many methods are used to improve the ability of geometric robustness. Nevertheless, no watermarking design is perfect that it can resist all geometric attacks. What we can do is to evaluate the practicability of watermarking system by contrast its capability and costs.This thesis mainly studies the problem that the geometric attacks often result in the failure of the digital watermark detection. A geometrically robust image watermarking algorithm is proposed by combining the fundamental principle and common requirement of digital watermark, using the advantages of Harmonic-Fourier moments in image description and making full use of the predominance of Harmonic-Fourier magnitude moments against geometric attacks. And two kinds of significant watermark method are designed in this thesis, one is significant watermark and the other is insignificant watermark. The watermark studied in this paper is invisible, robust image watermark; the detecting measure belongs to private watermark.In order to realize simplicity of extracting watermark, the algorithm adopts the method of embedding watermarking information in space field and extracting it in frequency field, which breaks the restriction of the embedding and extracting in conventional watermark design. When the carrier image suffers geometric attacks, conventional watermarking algorithm needs to make the carrier image reverse affine transformation. Whereas, the algorithm in this thesis needn't to restore the attacked images to original size, which greatly reduces the workload of watermark detecting and simplifies the algorithm.Experiments show that the algorithm has good rotational invariance and watermark has fine rubustness to scaling, JPEG compression and other noise attacks.