基于特征点和密度聚类的图像复制-粘贴篡改盲鉴别算法

Blind identification of image copy⁃move forgery based on feature points and dense clustering

针对数字图像复制-粘贴篡改盲鉴别方法存在的时间复杂度高、篡改定位不够精确等问题,提出了一种基于二进制描述符和密度聚类的数字图像复制-粘贴篡改的盲鉴别算法。该方法使用二进制描述符AKAZE提取特征点并描述特征,以降低特征计算的时间复杂度;使用基于密度的聚类算法DBSCAN去除错误匹配;最后通过仿射变换整幅图像对篡改区域进行定位。通过不同数据集上的实验结果表明:本文算法准确率最高可达到95%以上,误检率基本在10%以下,有效降低了时间复杂度,在可靠性和篡改定位准确率方面具有更优越的性能。

Image manipulation forensics technology has become a hot research topic in the field of image processing.Copy-move forgery detection is one of the most important and popular digital image forensics technologies.In order to solve the problems of high time complexity and inaccurate location of tampered regions in digital image copy-move forgery,a method of copy-paste forgery detection based on binary descriptor and density clustering algorithm is proposed.First,the binary descriptor AKAZE is used to extract feature points and describe features to reduce time complexity of feature calculation.Second,the density-based clustering algorithm DBSCAN is applied to remove false matches. Finally, the tamperedarea is located by affine transformation of the whole image. The experimental results on different datasetsshow that using the proposed algorithm, the highest localization accuracy exceeds 95%, and the falsedetection rate is basically less than 10%. It can effectively reduce the time complexity and show betterperformance in reliability and accuracy of forgery location.