The Study of Copyright Protection and Content Authentication for Digital Images

• Main Authors: Wei-Che Chen, 陳煒喆
• Other Authors: Ming-Shi Wang
• Format: Others
• Language: en_US
• Published: 2008
• Online Access: http://ndltd.ncl.edu.tw/handle/13351869600315431735

博士 === 國立成功大學 === 工程科學系碩博士班 === 96 === This dissertation addresses copyright protection and content authentication for digital images. In this context, we first propose a copyright protection scheme based on visual cryptography (VC) technique. Unlike most of conventional copyright protection schemes embedding a watermark into a host image directly that inevitable causes some permanent distortion, the proposed scheme conceals a binary watermark to two shares, according to the features of the host image and a predefined VC codebook, without modifying any pixel value of the host image. Therefore, the visual quality of the host image can be completely retained. In addition, the proposed scheme can use a binary image of arbitrary size as the watermark regardless of the size of host images. It is shown that the hidden watermark can be successfully revealed by the proposed scheme, even if the visual quality of protected images has undergone severe attacks, e.g., JPEG compression, noise addition and filtering. To further save the extra storage space required to store the codebook, another VC-based copyright protection scheme that conceals the watermark to two shares without resorting to any codebook is proposed. The enhanced scheme provides stronger robustness against several image processing operations when compared with the existing copyright protection schemes. Moreover, the experimental results demonstrate that it is difficult for malicious attackers to break the security of the proposed scheme. To ensure the integrity and authenticity of digital images, a wide variety of block-wise authentication schemes have been proposed in the literature. An effective authentication scheme should provide sufficient robustness to thwart any possible attacks with accurate tamper localization accuracy. However, most of the existing schemes either fail to address this issue or sacrifice tamper localization accuracy to withstand counterfeiting attacks, such as vector quantization (VQ) attack. For this reason, a novel block-wise authentication scheme based on fuzzy c-means (FCM) clustering is proposed for securely verifying the integrity of digital images and for identifying tampered image blocks. The proposed scheme successfully utilizes the FCM clustering technique to solve the security problem related to independent block-wise authentication schemes and at the same time preserves superior tamper localization accuracy. Furthermore, an image authentication scheme with tamper recovery capability is desirable, as it will save additional bandwidth to retransmit the original images if the authenticator can directly remedy the tampered ones. Accordingly, we propose a novel block-wise color image authentication scheme, which can not only detect and determine the locations where the tampering occurred, but also approximately recover the tampered blocks by the recovery data hidden in other blocks. To verify the validity of the proposed scheme, the tamper localization capability and tamper recovery capability of the scheme under different counterfeiting attacks are studied and quantified in this dissertation.