基于SIFT的图像复制遮盖篡改检测技术

Detection of copy-move image forgeries based on SIFT

针对现有的图像复制遮盖篡改检测算法需要某些先验信息、对后处理操作失效且计算量大等问题,借鉴图像匹配技术中的SIFT(scale invariant feature transform)特征匹配算法,首次提出将其用于检测复制遮盖的篡改操作。该方案首先计算出待鉴定图像的所有特征向量,然后进行特征向量集的划分与匹配。待鉴定图像的匹配点对用点与点之间连线标记。如果图像被复制篡改过,所标记的线段将明显集中于某两个区域之间。实验表明,SIFT应用于复制遮盖及各种后处理篡改操作,如旋转、缩放、亮度调整等,都有很好的检测和定位效果。

In view of the problem that the existing detection algorithms of the copy-move image forgeries have to rely on some premises,invalidation to post processing and have heavy computation,a new detection method using scale invariant feature transform（SIFT） for image registration was proposed. The scheme calculated all the eigenvestors of the tested image and performed division and matching operations on the set of eigenvestors. All the matching points were linked with lines located between the two regions obviously in a copy-move falsified image. Experimental results show a good performance for detecting and locating the copy-move image blocks in post processings such as rotation,scaling and intensity adjustment.