Digital Watermarking Technique Based on Self Similarity of A Image

• Main Authors: Hao-Kuan Tso, 左豪官
• Other Authors: Der-Chyuan Lou
• Format: Others
• Language: zh-TW
• Published: 2006
• Online Access: http://ndltd.ncl.edu.tw/handle/33597990523611063121

博士 === 國防大學中正理工學院 === 國防科學研究所 === 94 === Digital multimedia can be duplicated and distributed easily nowadays due to the rapidly development of multimedia technology and the Internet. However, the above behavior has severely violated the right of the legal owner already. Digital watermarking technique is an effective method for protecting the right of the legal owner. If the dispute happens, the embedded digital watermark can be extracted to verify who the legal owner is. To protect the copyright of digital multimedia, three digital watermarking methods (based on compression concept, self-similarity characteristic, visual cryptography technique and low-frequency characteristic) are proposed in the dissertation subject to the characteristics of imperceptivity and robustness. There is a common characteristic between digital watermarking and compression techniques, that is, after image compressed and embedding a watermark into an image, the image quality can not reduce. According to the reason, the compression concept can be applied to watermark embedding. The first method is to propose a digital watermarking method using compression-based concept and coefficient similarity characteristic. The proposed method is not only to keep image quality, but also is robust to image processing attacks. Many digital watermarking techniques based on self-similarity characteristic of fractal compression have been proposed. However, the drawback of fractal compression is time-consuming. Hence, the second method is to propose a fractal compression-based digital watermarking method using the characteristic of image self-similarity. Experimental results show that the proposed method is simple and robust to image processing attacks. Visual cryptography technique has been extensively applied to sharing secret information. This technique is simple and secure. Besides, it does not require complicated computation. Only if two images are overlaid, the secret information can be recovered. The third method is based on the concept of secret sharing and proposes a digital watermarking method using visual cryptography technique. By exclusive-or computation between two secret images, the copyright of digital image can be verified. Because low-frequency coefficients can survive after lossy compression, the last one method is based on the unchangeable characteristic of low frequency coefficients and designs a digital watermarking method using block-wise concept. Experimental results show that the proposed block-based method is more robust than the one based on pixel.