Rotation, scaling and translation invariant digital image watermarking

Abstract

Many digital image watermarking schemes have been proposed and are robust to common signal processing, such as compression and filtering. Geometric distortions, such as rotation, scaling, and translation (RST), are considered the most challenging attacks. This thesis is on RST invariant digital image watermarking. In this thesis, we introduce the fundamental theories and techniques necessary for RST invariant digital image watermarking, and briefly review the existing RST invariant image watermarking schemes. Then, we propose an image rectification scheme for RST invariant digital image watermarking. Rotation and scaling transformations in the spatial domain result in cyclical shifts in the log-polar domain, which is the log-polar mapping (LPM) of the magnitude of the Fourier spectrum of the image. We utilize this property to detect the rotation and scaling parameters by using a matching template and our new phase-only filtering method in the LPM domain. We employ the same strategy in the spatial domain to detect the translation parameters. This rectification scheme can detect RST parameters very accurately. (Abstract shortened by UMI.)