A neural network based lossless digital image watermarking in the spatial domain

Abstract

A lossless digital image watermarking algorithm based on neural network is proposed. The algorithm uses the neural network technique and the exclusive-or (XOR) operation to model the relationships among some randomly selected pixels with their neighborhoods, and to extract the features of the cover image instead of embedding watermark into it. It does not degrade the visual quality of the cover image, i.e., lossless watermarking. The algorithm may be used for automatic piracy detection as well as copyright demonstration by cooperating with a copyright authentication center. © Springer-Verlag Berlin Heidelberg 2005.