基于色彩制约耦合距离惩罚的图像篡改检测

Image copy-paste tampering detection algorithm based on color constrained coupling distance penalty model

当前较多图像篡改检测方法主要通过对图像特征间的距离进行测量来完成特征匹配,忽略了图像的色彩信息,导致检测结果中存在较多的误检测和漏检测现象。对此,本文将色彩信息引入到图像特征匹配过程中,设计了一种采用色彩制约模型的篡改检测算法。利用Laplacian算子与Harris算子提取图像特征,并利用像素点的红(R)、绿(G)、蓝(B)三原色信息,结合特征描述符建立色彩制约模型,对特征点间的色彩信息进行度量,再借助该度量值与特征点间的距离测量值共同完成图像特征匹配,充分剔除误匹配现象,有效提高匹配准确度。该算法还根据特征点间距离方差构造距离惩罚模型,对匹配后的图像特征进行聚类,准确识别篡改内容。通过实验结果发现,与其他篡改检测算法相比,本文算法不仅对伪造内容具备更高的检测准确度,而且对模糊及旋转等内容操作也具有更好的适应性。

At present,many image tampering detection methods mainly measure the distance between image features to complete feature matching,ignoring the color information of the image,resulting in more false detection and missed detection in the detection results.In this paper,color information is introduced into the process of image feature matching,and a tamper detection algorithm based on color constraint model is designed.The Laplacian operator and Harris operator are utilized to extract the image features.The R,G and B primary color information of the pixels,as well as the image feature descriptor are adopted to form a color restriction model for measuring the color information between feature points.Then the image feature matching is accomplished by using the distance measurement between the measure value and the feature points to fully eliminate the mismatch phenomenon and effectively improve the matching accuracy.Additionally,the distance penalty model is constructed according to the distance variance between feature points to cluster the matched image features for identifying the tampered content accurately.The experimental results show that compared with other tamper detection algorithms,the proposed algorithm not only has higher detection accuracy for forgery content,but also has better adaptability for content operations such as blur and rotation.