基于点匹配的区域拷贝篡改检测  

Detection of Image Region Forgery Based on Point Matching

作　　者：朱珏钰[1] 

机构地区：[1]湖南第一师范学院信息科学与工程系,湖南长沙410205

出　　处：《计算技术与自动化》2014年第2期97-100,共4页Computing Technology and Automation

基　　金：国家自然科学基金资助项目(60973113);院级一般科研项目(XYS10N05)

摘　　要：图像区域拷贝是一种常见的数字图像篡改技术,目前的大部分数字图像区域拷贝取证技术未考虑旋转和缩放因素。提出一种新的基于点匹配的图像区域篡改检测算法。首先利用尺度不变旋转变换(SIFT)寻找图像中的关键点,使用主成分分析法(PCA)对关键点进行降维描述,然后利用关键点特征向量的相似度寻找关键相似点。实验表明,该算法不但能够较精确地定位出复制和粘贴的图像伪造区域,还能有效抵抗噪声污染、有损压缩以及旋转等攻击,并有效地减少运算量,提高了检测效率。Copy move is a common digital image manipulation technique.Attacks such as rotation and scaling are not considered in most copy move image forensic methods.This paper proposes a novel region forgery detection algorithm based on point matching.Firstly,we extract feature points with Scale invariant Feature Transform （SIFT）,and describe the points by Principal Component Analysis （PCA）.Owing to the similarity between the pasted region and the copied region,we find out all possible forgeries by seeking for similar point pairs by using their descriptors.The experimental results demonstrate that our proposed algorithm performs not only robustly in terms of additive noise,lossy JPEG compression but also effectively in rotation attack,comparing with the algorithm based on block matching.Furthermore,our algorithm has reduced the calculation amount and improved the detection efficiency.

关 键 词：区域拷贝 点匹配 篡改检测 块匹配 关键点 

分 类 号：TP309.2[自动化与计算机技术—计算机系统结构]