摘要
目的提出一种将SIFT算法和CIE Lab颜色模型相结合的方法来检测复制-移动篡改图像,解决传统SIFT(Scale-invariant feature transform)算法无法应用颜色特征进行篡改图像检测所导致的特征关键点的错误匹配问题,提高篡改图像检测的准确度.方法分别提取图像的SIFT特征与Lab颜色特征;使用KNN(K-Nearest Neighbor)算法对提取的特征进行分类匹配,排除异常特征值.结果笔者所提方法与以往的SIFT算法相比较,其错误匹配个数明显下降,降低了时间复杂度,提高了检测准确率,对图像篡改部分的平移、缩放和旋转操作都具有较强的鲁棒性,这三种操作对应的F1值分别可达86.8%,88.4%,88.5%.结论 SIFT算法和CIE Lab颜色模型提取的特征能够较好地满足检测复制-移动篡改图像的要求,颜色信息能够有效地改善特征匹配效果,KNN算法能够成功地排除异常匹配点.
A method combining SIFT algorithm and CIE Lab color model is proposed to detect the copy-move forgery image,it solves the problem of wrong matching of feature key points caused by the traditional SIFT(scale-invariant feature transform)algorithm can not apply color feature to image tampering detection and improves the accuracy of tampering image detection.SIFT features and lab color features of images are extracted successively;the KNN(K-Nearest Neighbor)algorithm is used to classify the extracted features and remove abnormal features.The method proposed by the author is compared with SIFT algorithm in the past,the number of mismatches is significantly reduced,the time complexity is reduced and the detection accuracy is improved.The presented method is robust to the translation,scaling and rotation operations of image tampering,and the corresponding F1 values of the three operations can reach 86.8%,88.4%and 88.5%,respectively.The features extracted by SIFT algorithm and CIE Lab color model can meet the requirements of detecting forgery images well,the color information can improve the matching results effectively,and the KNN algorithm can exclude the abnormal matching points successfully.
作者
宋凯
覃圣淋
SONG Kai;QIN Shenglin(School of Information Science and Engineering,Shenyang Ligong University,Shenyang,China,11015)
出处
《沈阳建筑大学学报(自然科学版)》
CAS
CSCD
北大核心
2020年第5期932-938,共7页
Journal of Shenyang Jianzhu University:Natural Science
基金
国家自然科学基金项目(61672360)
辽宁省自然科学指导计划项目(2019-ZD-0260)。