具有几何约束匹配的SIFT图像盲检测研究  

Blind Detection of SIFT Images with Geometric Constraint Matching

作　　者：张露[1] 林开司[1] 张志宏[1] 

机构地区：[1]铜陵职业技术学院,安徽铜陵244000

出　　处：《重庆科技学院学报（自然科学版）》2017年第5期108-111,共4页Journal of Chongqing University of Science and Technology：Natural Sciences Edition

基　　金：安徽高等学校自然科学基金重点项目"数字图像盲取证技术研究与应用"(KJ2016A712)

摘　　要：复制粘贴是常见的图像篡改手段之一。SIFT是常用的一种匹配算法,也是一种有效的图像复制粘贴篡改检测方法。该算法只采用欧式距离作为度量图像复制区域和粘贴区域特征点向量间的相似性,并未加入其他的匹配约束,由此会产生特征点的误匹配;因此,对得到的匹配点对采用方向一致性几何约束,剔除误匹配点对,提高匹配的正确率和算法后续几何评估中RANSAC算法的执行效率。最后,分别对缩放、旋转,以及在篡改图像中添加噪声等情况进行实验。实验证明改进算法对上述情况的性能提升明显。Copy-paste is one of the common methods of image tampering.SIFT is a kind of common matching algorithm,and it is also an effective method for image copy-paste tampering detection.In this algorithm,the Euclidean distance is used as a measure of the similarity between the image copy region and the feature point vector of the paste region.There are no other matching constraints,which will result in the error matching of feature points.So direction consistency geometric constraint of the matching points is obtained.The false matching points are eliminated.The matching accuracy of the algorithm and the efficiency of RANSAC algorithm evaluation are improved.Finally,the scaling,rotation,and adding noise in the tampered images are tested.The experimental results show that the improved algorithm improves the performance of the above situations mentioned.

关 键 词：篡改 盲取证 几何约束 匹配 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]