基于特征域和Zernike矩的多倍零水印版权认证方案  

Multiple Zero-watermarking Authentication Scheme Using Feature Region and Zernike Moment

作　　者：高光勇[1,2] 蒋国平[3] 

机构地区：[1]南京邮电大学计算机学院,江苏南京210023 [2]九江学院信息科学与技术学院,江西九江332005 [3]南京邮电大学自动化学院,苏南京210023

出　　处：《南京邮电大学学报（自然科学版）》2012年第6期63-68,共6页Journal of Nanjing University of Posts and Telecommunications：Natural Science Edition

基　　金：国家自然科学基金(60874091);江苏省"六大人才高峰"高层次人才基金(SJ209006);高等学校博士点基金(20103223110003);江苏省自然科学基金(BK2010526);南京邮电大学引进人才项目(NY209021);江西省教育厅科技项目(GJJ12614);江苏省高校研究生科研创新计划(CXZZ11_0400)资助项目

摘　　要：利用特征域和Zernike矩,提出一种多倍零水印版权认证方案。该方案采用Harris角点检测去提取载体图像的特征点,然后将这些特征点扩展成半径为R的特征域。特征域的Zernike矩不变量被用来构造特征向量,并且将这些特征向量二值化以形成多倍零水印。在水印检测阶段,通过正确检测出的水印数量与阈值的比较来证明检测图像的版权属性。研究结果表明提出的方法能拒绝各种攻击,包括模糊、噪声、中值滤波、JPEG压缩、旋转、缩放、剪切及Stirmark RBA攻击。与同类算法相比,文中所提方法具有更好的性能。In this paper, based on feature region and Zemike moment, a multiple zero-watermarking copy- right authentication scheme is proposed. This scheme utilizes Harris comer detection to extract the feature points of cover image, then the feature regions are generated by expanding the feature points to the disk regions with the radius of R. The Zemike moment invariants of feature regions are calculated to construct feature vectors, which are then binarized to form multiple watermarking. The process of watermarking de- tection proves the ownership of copyright by the comparison between the number of correctly detected wa- termarking and a threshold. Simulations demonstrate that the proposed scheme is resistant to various at- tacks such as blurring, additive noising, median filtering, JPEG compression, rotation, scaling, cropping and Stirmark RBA. Compared with the existing algorithms, the developed method has better performance.

关 键 词：多倍零水印 特征域 ZERNIKE矩 鲁棒性 

分 类 号：TP309.2[自动化与计算机技术—计算机系统结构]