数字图像多区域复制粘贴盲检测方法  

Digital Image Multi-area Copy-move Blind Detection Method

作　　者：魏佳[1] 曹记东 李军[1] WEI Jia;CAO Ji-dong;LI Jun(School of Mathematics and Computer Science,Shaanxi University of Technology,Hanzhong 723001,China)

机构地区：[1]陕西理工大学数学与计算机科学学院,汉中723001

出　　处：《科学技术与工程》2019年第5期167-171,共5页Science Technology and Engineering

基　　金：陕西省教育厅科研专项(14JK1144);秦巴山区生物资源综合开发协同创新中心自然科学项目(QBXT-Z(P)-15-20);陕西理工大学科研项目(SLG1807)资助

摘　　要：随着网络技术和图像处理技术的发展,篡改图像变得越来越容易。特别是图像文件内部的复制粘贴篡改,更不容易察觉。如何快速有效的盲检测出复制粘贴区域,成为一个急需解决的问题。现有的盲检测技术对单区域复制粘贴盲检测效果较好;但对多区域复制粘贴的漏检率较高。提出一种基于特征点和k-Rg2NN算法的数字图像多区域复制粘贴盲检测方法;该方法首先对图像进行SURF特征点提取,生成特征描述子,采用提出的k-Rg2NN算法对描述子进行匹配,通过RANSAC算法计算图像中源区域与目标区域之间的单应变换,准确定位复制粘贴区域。实验结果表明,该方法能有效检测多区复制粘贴,达到较高的准确率。With the development of network and image processing technologies, forging image become so easy. It is very hard to detecting in-file copy-move forgery, especially. Quickly and effectively detecting copy-move area become a problem that need the solution urgently. Current methods perform well for single-area copy-move, but misdetection rate is high for multi-area. A digital image multi-area copy-move blind detection method was presented based on feature and k -Rg2NN algorithm. Image SURF feature points are extracted first, feature descriptors are generated, and they are matched by k -Rg2NN algorithm presented, homographic transformation between source area and target area are computed by RANSAC algorithm, copy-move areas are located accurately. The experimental results show that this method can efficiently detect multi-area copy-move, and achieve high accuracy.

关 键 词：数字图像 多区域 复制粘贴伪造 盲检测 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]