基于可视加密和特征匹配的鲁棒彩色图像水印  

Robust color image watermarking based on visual cryptography and feature matching

作　　者：颜斌[1] 王亚菲[1] 宋凌云[1] 崔鑫 

机构地区：[1]山东科技大学电子通信与物理学院,山东青岛266590 [2]海南省气象信息中心,海口570203

出　　处：《计算机应用》2015年第A01期53-56,73,共5页journal of Computer Applications

基　　金：国家自然科学基金资助项目(61272432);青岛市科技计划项目(12-1-4-6-(10)-jch);山东省自然科学基金资助项目(ZR2014JL044);山东科技大学研究生创新基金资助项目(YC140340)

摘　　要：针对彩色图像和彩色水印的抗几何攻击鲁棒水印问题,提出一种基于可视加密的算法。该算法利用空域统计相关性,提取载体图像上的鲁棒特征,将其与彩色水印通过彩色可视加密绑定。为抗旋转攻击,提出利用图像的局部结构,提取图像的尺度不变特征点,使用特征点匹配来校正几何攻击的影响。实验结果表明,该算法对于加噪、滤波、JPEG压缩和旋转等攻击具有较强的鲁棒性,特别是在任意角度的旋转攻击下,提取的水印和原始水印的归一化相关值都在95%以上,可应用于医学图像和科学数据图像的保护和认证。Aiming at the design of geometric attack robust watermarking system for color image and color watermark, a watermarking system based on visual cryptography was proposed. The robust features of the image were extracted where the spatial correlation were fully utilized. These features were bounded with the watermark by visual cryptography. In addition, the local structure of the host image was analyzed and the Scale Invariant Feature Transform （ SIFT） was used to match the local features between the attacked image and the original image. The experimental results show that this algorithm is robust to most common attacks, such as noise addition, image filtering （ blurring） , JPEG compression and rotation. In addition, for rotation attack with arbitrary rotation angle, the normalized correlation between the extracted watermark and the original watermark is above 95%. This algorithm is applicable in applications such as medical image and scientific image protection, where modifications of the host images are usually not allowed.

关 键 词：彩色数字水印 可视加密 空域统计相关性 几何攻击 尺度不变特征变换 

分 类 号：TP309.7[自动化与计算机技术—计算机系统结构]