基于对抗颜色空间梯度SURF特征匹配复制黏贴篡改检测  

SURF replication adhesive tamper detection based on anti-color spatial gradient

作　　者：秦铭谦[1] 曾青松[1] 

机构地区：[1]广州番禺职业技术学院

出　　处：《电子测量与仪器学报》2018年第10期147-153,共7页Journal of Electronic Measurement and Instrumentation

基　　金：广东省自然科学基金(2015A030313807);广州市属高校科研项目(1201610059);广州番禺职业技术学院“十三五”科研项目(2016X002)资助

摘　　要：为了提高篡改检测方法对复制区域几何变换的鲁棒性,以及解决常规匹配算法带来的误匹配问题,提出了一种基于对抗颜色空间梯度SURF特征匹配复制黏贴篡改检测方法,首先,将完成图像对抗颜色空间梯度的获取,采用梯度值替代灰度值完成SURF的128维特征描述子提取。然后,采用凝聚层次聚类算法对匹配特征关键点进行聚类处理,采用仿射变换估计对聚类后匹配关键点进行估计。最后,采用RANSAC算法优化估计结果。通过实验验证,该方法能准确实现特征点的精确匹配,并能准确定位复制粘贴区域,且对篡改区域的几何变换具有很强的鲁棒性。In order to improve the robustness of the detection method to the geometric transformation of the replica area, and solve the problem of mismatch caused by the conventional matching algorithm, this paper proposes a space based on the counter color gradient SURF feature matching copy paste tamper detection method. First of all, it will complete the gradient image against color space to obtain, using gradient value instead of grey value 128 D SURF feature descriptor extraction. Then, the aggregation hierarchical clustering algorithm is used to cluster the key points of matching features, and affine transformation estimation is used to estimate the key points of matching after clustering. Finally, RANSAC algorithm is adopted to optimize the estimation results. Experimental results show that the proposed method can match feature points and locate copy-move area accurately, it has strong robustness to the geometric transformation of the tampered area.

关 键 词：同图复制粘贴篡改 SURF特征 凝聚层次聚类算法 仿射变换估计 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]