Abstract
Many image watermarking schemes against geometric attacks have been proposed. Rotation, scaling, translation (RST) invariant image watermarking based on log-polar mapping (LPM) and Radon transform is one of them. In this paper, we analyze the embedding and detection process of the above-mentioned method. We focus on how to select the watermark data, how to embed the watermark into the original image, and how to implement approximative inverse log-polar mapping (ILPM) during the embedding process. We analyze watermark detection when a watermarked image is scaled or rotated, and propose to use the rotation rectification to battle rotation. Three detection methods are compared.