基于小波和奇异值分解的图像复制伪造区域检测  

Detection of Copy Forgery Regions in the Image Based on Wavelet and Singular Value Decomposition

作　　者：吴琼[1] 李国辉[1] 孙韶杰[1] 涂丹[1] 

机构地区：[1]国防科学技术大学信息系统与管理学院,湖南长沙410073

出　　处：《小型微型计算机系统》2008年第4期730-733,共4页Journal of Chinese Computer Systems

摘　　要：针对数字图像取证中一类常见的复制粘贴图像伪造,本文提出了一种基于小波变换和奇异值分解的检测算法.该算法利用小波变换提取图像的低频分量,对低频分量分块提取奇异值特征,这种特征描述形式对图像羽化或边缘模糊等处理具有鲁棒性.然后将特征矢量进行按行字典排序,并且配合图像块的偏移位置信息,进行图像复制伪造区域的检测和定位.实验表明该算法能够较精确地定位出复制和粘贴的图像伪造区域,并有效地减少了运算量,提高了检测效率.Aiming at a common copy-paste forgery in the research field of digital image forensics, this paper proposes a detection approach based on wavelet transform and singular value decomposition. Firstly, DWT is applied to the image, and SVD is used on fixed-size Mocks of low-frequency component in wavelet sub-band to yield a reduced dimension representation. The representation is robust to minor variations in the image due to edge blurring. The SV vectors are then lexicographically sorted, and copy forgery regions are localized by detecting for all neighborhood vectors. The experiments demonstrate that the proposed approach can not only localize the copy forgery regions accurately, but also reduce the amount of computation and improve the detection efficiency.

关 键 词：图像取证 小波变换 奇异值分解 复制伪造区域 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]