中文摘要：

针对遭遇了复制-粘贴篡改的图像,提出了一种基于点匹配的重复区域检测方法.在经过特征点提取与匹配之后,提出简单有效的聚类方法对匹配点进行聚类,然后采用RANSAC算法进行类间几何变换估计,据此得到原篡改图像的几何变换副本,最后提出基于像素相关性的区域生长方法来描述重复区域.仿真实验表明提出的聚类方法能够克服传统方法的不足,重复区域描述方法在描述效果和时间效率上,都优于现下主流方法.

英文摘要：

Aiming at the images which suffered copy-move tampering,a method of duplicated region detection based on point-matching is proposed.After the feature extraction and matching step,a simple but effective clustering method is presented to classify the matched points.Then the geometric transformation between the classes is estimated by the use of RANSAC algorithm.And the counterpart of the original tampered image is obtained according to the geometric transformation.Finally,a region growing algorithm based on the correlation of pixels is proposed to describe duplicated regions.Simulation results show that the presented clustering method can make up the flaw of traditional one and the proposed duplicated region description algorithm is superior to the mainstream approach both in property and efficiency aspects.