基于神经网络的数字水印技术研究-atlantispress.pdfVIP

International Conference on Advanced Information and Communication Technology for Education (ICAICTE 2014) The research of digital watermarking technology based on Neural Network Liling Zhang Dept of Computer Science and Engineering, Hubei University of Education, Wuhan, China,430205 hu_tu@ Abstract. In the paper a kind of hyperchaos neural network is creatively used to realize the encryption of digital images. By selecting a reasonable initial value for the hyperchaos neural network, an encrypted matrix can be attained to encrypt the digital watermark; meanwhile, the chosen 3-layer wave division technique is adopted to realize image’s encryption. The proposed method makes well use of the non-predictability of chaos system and the close dependence on initial value, as well as that the encrypted matrix cannot be got by illegitimate users. Thereto, the chosen 3-layer wave algorithm is also unknown to unauthorized users, which can powerfully protect the copyrights of digital images. The experiment results show that the method is sound and reasonable with great invisibility and robustness. Key Words: hyperchaos neural network system, initial value, encrypted matrix; Watermark 基于神经网络的数字水印技术研究 章丽玲 湖北第二师范学院计算机科学与工程系，武汉，中国，430205 hu_tu@ 摘要:神经网络系统是从人脑动力系统 目前混沌的研究被广泛地应用于通信系 中抽象出来的一类重要的人工神经计算模 统的同步、解密技术中，同时在信息的必威体育官网网址 型。将超混沌神经网络系统的特殊性引入到 通信方面也发挥着越来越重要的地位。将混 数字水印领域来，可有效实现数字图像的加 沌系统应用到图像加密与信息隐藏方面，可 密。本文通过选择超混沌神经网络的合理初 在知识产权保护等方面获得很广泛的应用。 始值，得到一个加密矩阵进行水印的加密， 因此，本文创造性地运用了超混沌神经网络， 并结合三层选择性小波分解技术实现图像的 实现了数字图像的加密。其中 s 加密矩阵通 加密。该方法通过混沌系统的不可预测性以 过选择超混沌神经