A public adaptive watermark algorithm for color images based on principal component analysis of generalized Hebb

Abstract

In this paper, a novel public digital color image watermark algorithm, by taking a color image as watermark, is put forward, by which the hidden layer carries much more information than it does by words or characters. The algorithm is based on principal component analysis of generalized Hebb adaptive algorithm in artificial neural network and to do adaptive quantitative coding for principal component coefficients according to the proportion of marginal or textural information of the watermark image. In addition, it adaptively adjusts the embedding depth according to the images features to ensure the invisibility of the watermark. By way of disparting and stochastic embedding into color image watermark, it increases the embedding robusticity of watermark. By utilizing block-concerned technique, it realizes public watermark image algorithm. It respectively applies in face-identifying watermark with abundant marginal information and figureprint watermark with rich textural information, and gains fine feedback.