摘要
为了克服当前较多图像篡改检测算法主要通过比较特征点的距离来完成伪造内容的识别,忽略了特征点邻域所含的信息量,导致检测结果中存在较多的漏检和误检等问题,采用图像的亮度特征和信息量特征,设计了一种新的图像篡改检测算法。首先,引入Forstner算子,计算图像像素点的Robert梯度,从图像中精确获取特征点;其次,在图像特征的邻域中,通过均值模型来计算图像的亮度特征,将其与像素点的灰度差异特征相结合,以构造健壮的特征向量;再次,采用互相关函数来计算图像特征的关联度,采用信息熵来评估图像特征邻域所含信息量;并以图像特征间的关联度与信息量特征为依据,对图像特征进行匹配;最后,利用图像特征的特征向量,获取匹配点间的距离值,实现匹配点的归类,获取检测结果。实验结果表明:与当下篡改检测算法相比,在多种几何内容变化下,所提算法具备更高的检测准确度,所含的漏检和误检信息最少。
In order to solve the problem as many missed detection and false detection phenomena in the detection results which induced by mainly comparing the distance of feature points to complete the identification of forged content and ignoring the information content of the neighborhood of feature points in current image tamper detection algorithms,a new image tamper detection algorithm based on the brightness feature and information feature of image was designed.First,the Forstner operator was introduced to calculate the Robert gradient of the pixels in the image for obtaining the feature points accurately.Then,the mean value model was used to calculate the brightness features of the image,which were combined with the gray-scale difference features of pixels to construct robust feature vectors in the neighborhood of image features.Finally,the correlation function was used to calculate the correlation degree of image features,and the information entropy was used to evaluate the information content of image feature neighborhood.In addtion,the image features were matched based on the correlation degree and information quantity characteristics of image features.Finally,the distance between matching points was obtained by using the feature vector of image features to classify the matching points for obtaining the detection results.Experimental results show that the proposed algorithm has a higher detection accuracy which contains the least missing and false detection information under various geometric content changes compared with the current tamper detection algorithm.
作者
王欣
徐平平
吴菲
WANG Xin;XU Ping-ping;WU Fei(Pujiang College of Nanjing University of Technology,Nanjing 211200,China;College of Information Science and Engineering,Southeast University,Nanjing 210096,China;National Mobile Communications Research Laboratory,Nanjing 210096,China)
出处
《科学技术与工程》
北大核心
2020年第33期13740-13746,共7页
Science Technology and Engineering
基金
国家科技重大专项(2018ZX03001008-002)
江苏省高校自然科学研究项目(19KJB520037)
江苏高校“青蓝工程”。
