基于曲率度量耦合仿射制约策略的图像复制-粘贴篡改检测算法  

Image copy-paste tampering detection algorithm based on curvature metric coupled affine constraint strategy

作　　者：卢淑萍[1] 肖随贵[2] Lu Shuping;Xiao Suigui(College of Big Data and Artificial Intelligence,Guangdong Polytechnic of Science and Technology,Zhuhai 519090,China;College of Mathematics and Computer Science,GanNan Normal University,Ganzhou 314000,China)

机构地区：[1]广东科学技术职业学院大数据与人工智能学院,珠海519090 [2]赣南师范大学数学与计算机科学学院,赣州314000

出　　处：《电子测量与仪器学报》2019年第9期129-136,共8页Journal of Electronic Measurement and Instrumentation

基　　金：国家自然科学基金(61370229);广东省自然科学基金(S2013010015178)资助项目

摘　　要：针对当前较多图像复制-粘贴篡改检测算法主要依靠对特征点间的距离进行度量来完成特征匹配,忽略了特征点间的仿射关系,使其在几何变换条件下的篡改检测检测正确性不高的问题,将特征点间的仿射关系引入到特征匹配过程中,提出基于曲率度量耦合仿射制约策略的图像复制-粘贴篡改检测算法。该算法主要是通过像素点曲率和特征点仿射关系来完成图像复制-粘贴篡改检测。首先,采用Sobel边缘检测方法提取图像的边缘轮廓,通过计算边缘轮廓上像素点的曲率值来获取图像特征。然后,通过计算特征点邻域中的Haar小波值,生成特征向量。利用特征向量构造特征点间的仿射关系模型,计算特征点间的仿射关系值,用于建立仿射制约策略,完成特征匹配。最后,借助于SURF算法完成特征点的集群,对复制-粘贴篡改区域进行定位,获取检测结果。实验结果显示,较当前的复制-粘贴篡改检测方法而言,所提算法具有更高的检测正确性与鲁棒性,能够更好地适应缩放、旋转等伪造内容的检测。In view of the fact that most current image copy-paste tampering detection algorithms mainly rely on measuring the distance between feature points to complete feature matching, but ignoring the affine relationship between feature points, which leads to the weakness of the algorithm in robustness and detection accuracy. In this paper, affine relationship between feature points is introduced into feature matching process, and a new method based on curvature and feature of pixel feature points is proposed. The point affine relation is used to complete the image copy-paste tamper detection algorithm. Firstly, Sobel edge detection method is used to extract the edge contour of the image, and the curvature value of the pixels on the edge contour is calculated to obtain the image features. Then, the feature vectors are generated by calculating the Haar wavelet values in the neighborhood of the feature points. The affine relation model between feature points is constructed by using eigen vectors, and the affine relation values between feature points are calculated to establish affine restriction strategies and complete feature matching. Finally, the cluster of feature points is completed by using Hough algorithm, and the copy-paste tampered area is located to obtain the detection results. The experimental results show that the proposed algorithm has better detection accuracy and robustness than the current copy-paste tampering detection method, and can better adapt to the detection of forged content such as scaling and rotation.

关 键 词：复制-粘贴篡改检测 曲率度量 Haar小波值 仿射关系模型 仿射制约策略 HOUGH算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TN01[自动化与计算机技术—计算机科学与技术]