一种基于JADE的数字图像盲水印算法  

An Algorithm of Digital Image Blind Watermarking Based on JADE

作　　者：李生[1] 朱苏阳[1] 赵煜[1] 马小虎[1] LI Sheng, ZHU Su-yang, ZHAO Yu, MA Xiao-hu （School of Computer Science and Technology, Soochow University, Suzhou 215006, China）

机构地区：[1]苏州大学计算机科学与技术学院,江苏苏州215006

出　　处：《电脑知识与技术》2011年第11期7723-7725,共3页Computer Knowledge and Technology

基　　金：江苏省大学生创新性实验计划项目（SG31505110）

摘　　要：为了提高数字水印的鲁棒性，并实现水印的盲提取，提出一种基于离散余弦变换（DCT），离散小波变换（DWT）和独立成分分析（ICA）数字水印算法。在水印嵌入时，将原始图像做一次DWT变换以提取低频子带，对该低频子带进行DCT变换并按水印大小分块，最后依据人类视觉敏感度自适应嵌入。在水印提取时，读取可能遭受攻击的水印图像，调JADE算法提取多个水印副本，选择其中相似度最接近的两个水印副本进行PcA融合，最终得到融合后的水印图像。实验结果证明了该方法对于JPEG压缩、噪声、旋转等具有很好的鲁棒性。An algorithm of digital watermarking based on JADE is proposed to improve the robustness of digital watermarking, and achieve the blind source separation. In watermark embedding procedure, we use DWT transform to obtain the low frequency sub-band, and then we transform it to the DCT domain and block it according to the size of the watermark, at last we choose the proper blocks to embed watermark. In watermark extracting procedure, we read the image which may be subject to attack, and calls JADE algorithm to extract multiple copies, select two copies which have the closest similarity to perform PCA fusion. Experimental results have demonstrated that the proposed approach is robust against common signal processing such as JPEG compression, noise and rotation.

关 键 词：数字水印 人类视觉敏感度 独立成分分析 主成分分析 联合近似对角化 

分 类 号：TP18[自动化与计算机技术—控制科学与工程;自动化与计算机技术—控制理论与控制工程]