基于量化相关性的JPEG图像盲取证  

Detecting Image Forgeries via JPEG Quantization Coherence Measurement

作　　者：吴首阳[1] 刘铭[1] 

机构地区：[1]中国科学技术大学电子工程与信息科学系,安徽合肥230027

出　　处：《计算机仿真》2010年第6期258-261,266,共5页Computer Simulation

基　　金："图像处理与图像通信"江苏省重点实验室开放研究课题(ZK207007)

摘　　要：针对真实图像检测问题,通过图像处理和编辑软件可以方便地修改数字照片,而数字图像真伪的盲取证技术正是为了解决检测中各种信任危机。通过研究JPEG压缩过程中的量化相关性特征,提出一种基于量化相关性测度的真伪图像盲检测方法,能够检测JPEG图像的真伪,并标定修改区域。该方法具有较高的灵敏度,可以对不同压缩参数的图像进行检测处理。实验结果表明,即使待检测图像经历过多次不同质量因子的JPEG压缩,方法同样具有有效性和鲁棒性。Blind digital image forensic is a technique to solve the trust crisis for digital pictures which might be doctored with powerful image processing software. A novel approach was proposed to detect doctored JPEG images and further locate the doctored area automatically. This approach was mainly accomplished by means of quantization coherence in JPEG compression, and the doctored region can be detected via test compressing again. The detection has high sensitivity and low computing cost. Experimental results show that this approach is effective and robust.

关 键 词：图像伪造 盲检测 压缩图像 量化效应 量化相关性 

分 类 号：TN911.73[电子电信—通信与信息系统]