Digital watermarking for image content authentication

ii tampering detection is enabled in the authentication process by presetting the noise filter size. Second, we study the unique characteristics of synthetic images and develop a novel watermarking scheme for synthetic image authentication. The proposed watermarking algorithm is fully compatible with the characteristics of synthetic images. With less pixels modified, the authentication system can still achieve pixel-wise tamper localization resolution. Moreover, we propose a new embedding strategy, which enables the capability of recovering the altered image content of the authentication system. Hence, not only can the authenticator localize the tampered area but also it is able to recover the removed content and identify the forged parts. In addition, in order to tackle the high image fidelity requirement in some special applications, we propose a framework for ROI-supporting watermarking systems, which can be applied to different watermark embedding schemes. Based on the framework, we present the non-ubiquitous watermarking schemes with ROI masking for natural images and synthetic images respectively. Unlike the common holistic watermarking schemes, the ROI-based schemes do not embed the watermark ubiquitously over the whole image but avoiding modifying the content of the specified important regions. Although there is no watermark embedded inside these regions, their integrity is still well protected as well as other image parts. The same tamper detection resolution is achieved both inside and outside the specified ROI regions.