Applications of a Neural Network to Estimate Watermark Embedding Strength

Digital watermarking technology has the advantages of easy implementation and capability of providing wide security services such as copyright protection, authentication and secret communication. In this paper, a digital watermarking scheme using neural network is proposed. Firstly, the original image is divided into some 8times8 blocks, and then neural networks according to different textural features and luminance of each block decide adaptively different embedding strengths. The watermark detection adopts correlation technology. Experimental results show that the proposed scheme has good imperceptibility and high robustness to common image processing operators.