摘要
分析了目前数字水印抗几何攻击的实现方法,对SIFT进行了改进。提出了特征点筛选算法,以增强提取特征点的鲁棒性,然后基于攻击前后图像的SIFT特征点匹配技术获取匹配点对,并基于匹配点对的尺度特征和SIFT中心距之比获得尺度失真参数以实现尺度失真的恢复。同时基于匹配点对的SIFT中心角度差确定旋转失真参数,进而完成旋转校正。实验结果表明,该算法能够有效进行水印图像几何失真参数的估计与几何失真的校正。
Implementation of current technologies in geometric robust image watermarking are analyzed, and modification to SIFT combined with selection of feature points are proposed, then SIFT feature points are detected both in distorted image and undistorted image, the matched points are searched, the scale and central distance features of the matched points are used to calculate the scale distortion parameter, then the scale distortion is restored. Meanwhile, The SIFT central angle of matched points is calculated to restore the rotated distortion. The experiments indicated that the algorism is efficient in geometric distortion correction.
出处
《计算机工程与设计》
CSCD
北大核心
2008年第12期3215-3217,共3页
Computer Engineering and Design
基金
江苏省自然科学基金项目（BK2006569）
长三角重大科技联合攻关基金项目（BE2004400）.
