摘要
针对图像的区域拷贝篡改提出了一种新的检测方法,即基于LPP-SIFT(Locality Preserving Projection-Scale Invariant Feature Transform)的检测算法.该算法利用SIFT提取图像的特征点,结合LPP特征映射,生成低维特征点描述子,然后依据向量角度比率得到匹配点对,匹配点对聚集的区域即为拷贝篡改区域.实验表明该算法能够有效地检测定位区域拷贝篡改图像,且对于拷贝区域的旋转、缩放以及图像的JPEG压缩、噪声添加等操作具有较强的鲁棒性.
This paper proposes a novel method based on LPP-SIFT（Locality Preserving Projection-Scale Invariant Feature Transform） for image copy-forgery detection.First,some steps of SIFT algorithm are used to extract key points;second,LPP is used to obtain low-dimensional feature descriptors,and then key points can be matched by the ratio of vector angel.If mass matching key points exist in a region,the region can be judged to have undergone copy-forged operation.Experiments demonstrate that the method can effectively detect forged images and locate the forged area,and this method is robust to rotation,scaling,JPEG compression and noise addition.
出处
《郑州大学学报（工学版）》
CAS
北大核心
2012年第6期15-18,共4页
Journal of Zhengzhou University（Engineering Science）
基金
河南省重大科技攻关计划项目(092101210101)
