基于模糊聚类的小波域半脆弱水印算法  

Semi-fragile watermarking algorithm based on fuzzy clustering in the wavelet transform domain

作　　者：康志伟[1] 肖娜[1] 何怡刚[2] 刘劲[1] 杨高波[1] 

机构地区：[1]湖南大学计算机与通信学院,长沙410082 [2]湖南大学电气与信息工程学院,长沙410082

出　　处：《仪器仪表学报》2008年第9期1851-1855,共5页Chinese Journal of Scientific Instrument

基　　金：国家自然科学基金(60702065);湖南省自然科学基金(06JJ50114);湘潭市科技计划重点项目(ZJ20071008)资助

摘　　要：将模糊C-均值聚类(FCM)与人类视觉系统的掩蔽性相结合,并利用小波系数奇异值的稳定性,提出了一种小波域半脆弱水印算法。该算法在对宿主图像分块并进行小波分解后,以各子块的低频背景与高频纹理为特征向量,利用FCM聚类方法进行特征分析,得到掩蔽度由弱到强的多个聚类中心,从而确定各子块的嵌入强度,自适应地将水印嵌入到低频小波系数的奇异值中。实验结果表明,该算法具有较好的隐蔽性、抗JPEG压缩性和对恶意攻击的敏感性,并可进行精确定位。This paper introduces a semi-fragile image watermarking algorithm in the wavelet transform domain. The algorithm is implemented by combining the fuzzy clustering of c-mean（FCM） with the imperceptible characteristic of a human visual system and exploiting the stability of the singular value of wavelet coefficients. First, the host image is divided into blocks, and each block is then decomposed through wavelet transformation. The low-frequency background and high-frequency textures of each block are used as a feature vector, and by fuzzy c-means clustering （ FCM）, the characteristics of the image are analyzed. Several cluster centers arranged from strong to weak according to their masking capability are determined. Finally, the watermarking is adaptively embedded to the singular value of wavelet coefficients according to the degree of strongness of the masking centers. Experimental results indicate that the proposed algorithm has good imperceptibility, robustness to JPEG compression, fragility to malicious attacks and accurate tamper localization.

关 键 词：半脆弱水印 模糊聚类 人类视觉系统 小波变换 

分 类 号：TP309.2[自动化与计算机技术—计算机系统结构]