神经网络的Contourlet域水印研究  

Watermarking research of Contourlet transform of ANFIS network

作　　者：文娟[1] 谭阳红[1] 陈建[1] 

机构地区：[1]湖南大学电气与信息工程学院,湖南长沙410012

出　　处：《计算机工程与设计》2010年第15期3459-3463,共5页Computer Engineering and Design

基　　金：国家自然科学基金项目(50277010);湖南省自然科学基金项目(07JJ6132;04FJ2003;05GK2005);教育部博士基金项目(20020532016);高等学校新世纪优秀人才支持计划基金项目(NCET-04-0767)

摘　　要：为了更好地保证数字水印的鲁棒性和安全性,提出了一种基于Contourlet变换分解和ANFIS网络结合的数字水印算法。宿主图像先进行Contourlet分解,将置乱后的水印嵌入到Contourlet分解的低频系数中。根据ANFIS网络能够很好的模拟出输入输出之间关系的特点,训练好逼近宿主图像和水印图像之间的关系的ANFIS网络。在接收端用训练好的ANFIS网络盲提取水印,实现了在Contourlet域中采用神经网络盲提取的水印算法。实验结果表明,该水印算法安全性较高,经过噪声干扰、滤波、压缩等攻击具有很好的鲁棒性。A digital watermarking algorithm is proposed based on ANFIS network and Contourlet transform,to guarantee its robustness and secutity.A watermark by scrambling is embedded into low-frequency coefficients of the host image through Contourlet transform.According to the characteristics of simulating a good relationship between input and output by ANFIS network,the ANFIS network is trained.Then the watermark is blind extracted by the trained ANFIS network in the receiver.So the blind watermark extraction algorithm is achieved based on ANFIS network and Contourlet transform.Experimental results show that the watermarking algorithm is a high security.And it has a good robustness through noise,filtering,compression and other attacks.

关 键 词：数字水印 CONTOURLET变换 水印置乱 自适应模糊网络 鲁棒性 

分 类 号：TN911.73[电子电信—通信与信息系统]