| 摘要: |
| 随着近些年成本低廉的高性能电子成像设备的不断普及和操作简单的数字图像编辑软件的广泛应用,人们制作一幅篡改图像已经变得越来越容易。这些技术使得人们很难察觉和辨识那些使用专业技术处理过的篡改图像的伪造痕迹,因而对包括新闻传播、司法取证、信息安全等诸多领域带来了严重的威胁,数字信息的安全性和可靠性也因此越来越受到国际社会的广泛关注。综上所述,开展针对数字图像篡改检测方法的研究有着极其重要的意义。本综述围绕数字图像篡改盲检测方法开展工作。首先,本文根据数字图像篡改检测方法所依赖的线索对篡改检测方法进行层次化分类,将图像篡改检测方法分为两个方面:基于成像内容及成像系统印记一致性的检测方法和基于篡改及JPEG重压缩痕迹的检测方法。然后,按照内容的来源和篡改操作所处的阶段,将以上两方面篡改检测方法进一步分为四个分组:基于成像内容一致性的检测方法、基于成像系统印记一致性的检测方法、基于篡改及其后处理痕迹的检测方法和基于JPEG重压缩痕迹的检测方法;又根据目前文献涉及话题的分布情况,再将四个分组细分为十二个分类:基于光照一致性的检测方法、基于特征提取与分类的检测方法、基于成像色差印记一致性的检测方法、基于自然模糊印记一致性的检测方法、基于成像系统噪声印记一致性的检测方法、基于彩色滤波阵列插值印记一致性的检测方法、基于几何变换及插值痕迹的检测方法、基于人为模糊痕迹的检测方法、基于中值滤波痕迹的检测方法、基于特征匹配的检测方法、基于对齐JPEG重压缩假设的检测方法和基于非对齐JPEG重压缩假设的检测方法。接着,本文梳理出每种分类的主干的思想脉络并对该类中重要的算法加以详尽分析和论述。除此以外,本文还对各类方法中典型的算法的性能做了比较,并归纳总结了在各种方法中常见的性能衡量标准和公开数据集,便于后续研究使用。最后,本文对各方法存在的问题加以总结,并对未来发展的趋势做出预测。希望此综述能够对数字取证有关的研究者提供研究文献的参考、研究方法上的启发和研究思路上的借鉴。 |
| 关键词: 图像篡改 盲检测 成像内容 成像系统 篡改痕迹 |
| DOI：10.19363/J.cnki.cn10-1380/tn.2022.05.05 |
| 投稿时间：2019-10-27修订日期：2019-10-27 |
| 基金项目:本课题得到国家重点研发计划课题(No.2019QY2202,No.2020AAA0140000)资助 |
|
| A Survey on Blind Detection of Tampered Digital Images |
| ZHANG Yixuan,ZHAO Xianfeng,CAO Yun |
| State Key Laboratory of Information Security, Institute of Information Engineering, Chinese Academy of Sciences, Beijing 100093, China;School of Cyber Security, University of Chinese Academy of Sciences, Beijing 100093, China |
| Abstract: |
| In recent years, with the popularity of cheap and advanced electronic imaging equipment and user-friendly image editing software, it is becoming easier for people to make tampered images. The traces of tampered images processed with professional techniques are hardly noticeable, which poses serious threats to many areas including news broadcasting, judicial forensics and information security, hence the security and reliability of digital information have been receiving more and more attention from all over the world. In summary, it is extremely important to carry out research on digital image tampering detection. This survey focuses on blind detection of tampered digital images. Firstly, according to the clues on which the digital image tampering detection method relies, the tampering detection method is hierarchically classified into two aspects:the one is consistency of imaging content and imprint of imaging system based method, the other is tampering trace and JPEG recompression trace based method. Then, according to the sources of the content and the stages of the tampering operation, the two aspects of tampering detection are further divided into four groups:consistency of image content based method, consistency of imprint of imaging system based method, tampering and its post processing based method and JPEG recompression based method. Based on the distribution of the topics in the current literature, the four groups can be further subdivided into 12 categories:consistency of illumination based method, feature extraction and classification based method, consistency of color difference imprint based method, consistency of natural blurring imprint based method, consistency of imaging noise imprint based method, consistency of color filter array interpolation imprint based method, geometric transformation and interpolation trace based method, artificial blurring trace based method, median filter trace based method, feature matching based method, aligned JPEG recompression hypothesis based method and non-aligned JPEG recompression hypothesis based method. What's more, we summarize the main ideas in each category, with describing and analyzing important algorithms in each category. In addition, we compare the performance of typical algorithms in each category and summarize evaluation metrics and public datasets in all categories, which will be helpful for subsequent research. Finally, we sum up the shortcomings and predict the trend in each category. We hope that this survey can act as a reference and inspiration and can provide ideas for future researchers. |
| Key words: image tampering blind detection imaging contents imaging equipment tampering traces |
