Robust image watermarking scheme with general regression neural network and FCM algorithm

Abstract

A robust image watermarking method based on general regression neural network (GRNN) and fuzzy c-mean clustering algorithm (FCM) is proposed. In order to keep the balance between robustness and imperceptible, it uses FCM to adaptively identify watermarking embedding locations and strength based on four characteristic parameters of human visual system. For good learning ability and fast train speed of GRNN, it trains a GRNN with the feature vector based on the local correlation of digital image. Then embed and extracted watermark signal with the help of the trained GRNN. In watermark extracting it does not need original image. Experimental results show that the proposed method has better performance than the similar method in countering common image process, such as Jpeg compression, noise adding, filtering and so on. © 2008 Springer-Verlag Berlin Heidelberg.