图像拼接篡改的自动色温距离分类检验方法  

Detection of Image Splicing Manipulation by Automated Classification of Color Temperature Distance

作　　者：孙鹏[1,2,3] 郎宇博 樊舒 沈喆[4,5] 彭思龙 刘磊[1] SUN Peng;LANG Yu-Bo;FAN Shu;SHEN Zhe;PENG Si-Long;LIU Lei(Criminal Investigation Police University of China,Shenyang110035;Institute of Automation,Chinese Academy of Sci-ences,Beijing 100190;National Engineering Laboratory ofEvidence Traceability Technology,Beijing 100038;Liaoning Shihua University,Fushun 113001;Shenyang Institute ofAutomation,Chinese Academy of Sciences,Shenyang 110016)

机构地区：[1]中国刑事警察学院,沈阳110035 [2]中国科学院自动化研究所,北京100190 [3]现场物证溯源技术国家工程实验室,北京100038 [4]辽宁石油化工大学,抚顺113001 [5]中国科学院沈阳自动化研究所,沈阳110016

出　　处：《自动化学报》2018年第7期1321-1332,共12页Acta Automatica Sinica

基　　金：国家自然科学基金(61307016);现场物证溯源技术国家工程实验室开放课题(2017NELKFKT09);中央高校基本科研业务费项目(D2017021)资助~~

摘　　要：拼接篡改是一类常见的图像伪造手段,现有取证方法难以实现图像中拼接篡改区域的自动检测与精确定位,导致拼接篡改伪造图像的取证长期依赖人工经验.基于图像中原始区域与拼接篡改区域所反映的光源色温的差异性,提出一种自动色温距离阈值分类的图像拼接篡改检测与定位方法.首先,变换待检验图像至YCb Cr色彩空间,并按照Grid-based方式结构化分解为大小的子图像块;然后,利用自动白平衡(Automatic white balance,AWB)中的白点检测原理对每一个子图像块进行色温估计,计算子图像块与参考区域之间的色温距离;最后,采用最大类间方差法自适应地求取色温距离分类的最佳阈值,对子图像块进行分类标注,实现了图像拼接篡改区域的自动检测与精确定位.实验表明,该方法能够实现图像拼接篡改区域的自动检测与定位,具有较高的量化检测精度.Splicing is a common types of tampering in image manipulation. As many authentication methods cannot detect and localize the manipulated area automatically in splicing images, authentication of splicing image has depended on human experience for a long time. In this paper, considering the inconsistency of color temperature between original area and splicing area, we propose an automated distance threshold classification method for splicing image detection and manipulation localization by color temperature estimation. At first, we transform suspicious image into YCb Cr color space and divide it into blocks with grid-based manner. Then, we estimate color temperature of each block using automatic white balance（AWB） theory, and calculate Euclidean distance between reference area and suspicious area. Finally, we localize the splicing area with an automated estimated optimal threshold of color temperature distance. Experiments indicate that our method can detect splicing images and localize splicing area effectively and automatically with a quantitative result.

关 键 词：拼接篡改 图像取证 色温估计 色温距离 自动阈值 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]