基于双JPEG压缩统计特性的图像篡改检测方法  

Detection Method of Digital Forensics Based on Double-JPEG Compression Statistical Characteristics

作　　者：岳红原[1] 黄坚[1] 

机构地区：[1]南京工程学院计算机工程学院,江苏南京211167

出　　处：《南京工程学院学报（自然科学版）》2009年第2期34-40,共7页Journal of Nanjing Institute of Technology(Natural Science Edition)

基　　金：南京工程学院科研基金项目(KXJ07041)

摘　　要：为有效识别和检测数字篡改图像,提出一种新的基于双JPEG压缩统计特性的检测方法.这种方法是通过检测双JPEG压缩图像的图像块先前的量化矩阵来实现.首先,定义在两次压缩时,出现在单独的DCT系数直方图的一些特性.然后,使用支持向量机来估计双压图像的初始的量化矩阵,有不同量化矩阵的图像块便是值得怀疑的区域.最后,基于大量不同的双压图像集,用一系列的实验来验证本方法的性能及可靠性.In order to identify and detect digital forensics effectively, this paper puts forward a new detection method based on double-JPEG compression statistical characteristics. The detection was conducted by detecting the previous quantification coefficients of double-JPEG compressed image pieces. To begin with, features of DCT histograms of particular coefficients of double compression were examined. Then support vector machines were used to estimate the original quantification coefficients from double-compressed images and the image areas with different quantification coefficients are of concern. Finally, a series of experiments was carried out to verify the performance and reliability of this method based on various double-JPEG compressed images.

关 键 词：统计特性 双JPEG压缩 支持向量机(SVM) 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]