基于改进Hu不变矩的图像篡改检测算法  

Image tamper detection algorithm based on improved Hu invariant moments

作　　者：闫旭 姜威[1] 贲晛烨[1] YAN Xu;JIANG Wei;BEN Xianye(College of In{ormation Science and Engineering, Shandong University, Jinan 250100, China)

机构地区：[1]山东大学信息科学与工程学院

出　　处：《光学技术》2018年第2期171-176,共6页Optical Technique

基　　金：国家自然科学基金项目(61571275)

摘　　要：提出了一种针对图像的复制-粘贴篡改进行检测的方法。算法以图像子带编码为基础,利用Hu不变矩和Zernike矩的不变量对篡改图像进行检测。对图像做子带编码处理,得到近似子带（LL）图像,对近似子带图像进行Hu不变矩和Zernike矩的特征提取,用来表征图像的信息,最后采用上述特征信息进行图像的相关匹配。实验中将所提算法与极余弦变换（PCT）算法以及图像Zernik矩算法进行比较,结果表明所提算法对图像的旋转、平移攻击具有更优质的鲁棒性,提高了检测准确率。A method for detection of copy-paste tampering is proposed.The algorithm which is based on the image sub-band coding uses the Hu invariant moments and the Zernike moments invariants to detect the tampered image.The method firstly makes sub-band coding processing for the image to get the approximate sub-band（LL）image,and then makes Hu invariant moments and Zernike moments to extract the features which can represent the image information based on the approximate sub-band image,and uses the above feature information to make image matching finally.The experiments make some comparison for the improved algorithm with the classical polar cosine transform（PCT）algorithm and the single image Zernike invariant moments algorithm.The results show that the proposed algorithm has better robustness for the image rotation,scale change and translation attacks,and it improves the detection accuracy.

关 键 词：子带编码 ZERNIKE矩 HU不变矩 近似子带 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]