基于TSVD的图像复制区域伪造检测算法  

Copy-region forgery detection based on TSVD in digital image forensics

作　　者：康晓兵[1] 魏生民[1] 

机构地区：[1]西北工业大学现代设计与集成制造技术教育部重点实验室,西安710072

出　　处：《计算机应用研究》2008年第12期3741-3743,共3页Application Research of Computers

基　　金：高等学校博士学科点专项科研基金资助项目(20050699032)

摘　　要：提出了一种有效的盲检测算法来识别图像复制区域伪造。该算法采用截尾奇异值分解(truncated sin-gular value decomposition,TSVD)变换来处理图像块数据,并对图像块进行相似性匹配检测。实验结果表明,本算法具有较强的检测能力,能够有效抵抗多种修饰操作,如JPEG有损压缩、高斯模糊、高斯白噪声污染等。This paper described an efficient and passive-blind approach to detect and identify the location of copy-region forgery and tampering in a digital image. The algorithm first partitioned an image into many small blocks with appropriate size, employed TSVD to transform and described these blocks on low-dimension representation, and then detected the similarity between these blocks. Experimental results also demonstrate the validity of the proposed approach on several forged images undergone retouching or modifications to conceal traces of forgery, such as lossy JPEG compression, Gaussian blur filter, Gaussian white noise contamination and so on.

关 键 词：数字图像取证 截尾奇异值分解 复制区域伪造 被动盲检测 

分 类 号：TP309.2[自动化与计算机技术—计算机系统结构]