基于SIFT和区域生长的图像复制粘贴盲检测算法

An Image Copy-move Blind Detection Algorithm Based on SIFT Descriptors and Region Growing

为了检测出经过缩放、旋转等操作的篡改区域,提出了一种基于尺度不变特征(SIFT)特征点和区域生长的复制篡改盲检测算法。该算法首先采用SIFT描述符提取图像的关键点,对关键点进行匹配;然后对匹配的关键点进行聚类,计算出每个聚类的中心;最后以聚类的中心为种子进行区域生长得出篡改区域。实验结果表明,该算法能较好的检测经过几何变换的篡改区域,可用于面向图像内容的真实性认证。

An image copy-move blind detection algorithm based on SIFT（ scale-invariant feature transform） de- scriptors and region growing is proposed in order to detect those areas which are tampered because of rotation and scaling. Firstly, the algorithm extracts the key points of the images with SIFT descriptors and matches the key points. Then the matched key points are clustered, and the center of each cluster is calculated. Finally, the tampe- red region is obtained after the growing of the region with the center of each cluster as the seed. The experiment re- sults show that the proposed algorithm is effective enough to test the tampered areas with geometric transformation so as to be used for image authenticity certification.