基于改进SIFT算法的图像复制粘贴盲检测方法  

Image Copy and Paste Blind Detection Method Based on Improved SIFT Algorithm

作　　者：林开司[1] 张露[1] 林开武 

机构地区：[1]铜陵职业技术学院电气工程系,安徽铜陵244000 [2]铜陵有色股份铜冠信息科技有限责任公司,安徽铜陵244000

出　　处：《重庆科技学院学报（自然科学版）》2016年第4期116-119,共4页Journal of Chongqing University of Science and Technology：Natural Sciences Edition

基　　金：安徽省高等学校自然科学研究重点项目"数字图像盲取证技术研究与应用"(KJ2016A712)

摘　　要：对高斯二阶差分(D^2OG)算子的产生过程和理论依据进行了分析,介绍了一种称为广义特征点匹配的新方法。通过仿真,先获得被检测图像的基于D^2OG算子的SIFT特征点,然后利用广义特征点匹配方法检测图像是否被篡改。对100幅篡改图像进行篡改检测,采用D^2OG算子得到的特征点与DOG算子得到的特征点用于检测图像复制粘贴的正确率接近;但用D^2OG算子检测效率更高,对图像检测量大、实时性要求高的检测任务有实用价值。The Gaussian second order difference （D^2OG） operator produce process and theoretical basis are analyzed and a new generalized feature points matching method is introduced. Through simulation, the image based on D^2OG operator SIFT feature points was obtained; then we could use the generalized feature point matching method to detect whether the image has been tampered. Finally, 100 tampered pictures were carried out the image tamper detection. Statistics shows that compared to the DOG operator, the feature points from DeOG operator is more close to copy and paste. But D^2OG operator is more efficient with larger amount of image detection, so it can be applied to high real- time demand situation.

关 键 词：高斯二阶差分 特征点匹配 篡改 复制 粘贴 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]