Neural-network-based zero-watermark scheme for digital images

Digital watermarking has been widely studied for the copyright protection of digital multimedia data. In traditional digital image watermarking techniques, a watermark signal is inserted into the host image, which generally introduces visual quality degradation. On the contrary, zero-watermark techniques extract some features from the host image and use them for watermark detection, instead of watermark embedding. In this method, the distortion problem to the host image due to watermark embedding is eliminated. The zero-watermark scheme proposed in this work employs neural networks for features extraction. It can be used for automatic piracy detection as well as copyright demonstration if it is associated with a copyright authentication center. (c) 2006 Society of Photo-Optical Instrumentation Engineers.