于局部相位量化的图像伪造检测算法  

Image forgery detection algorithm based on local phase quantization

作　　者：李伟征[1] 白光伟[1] 涂秀梅[2] LI Wei-zheng;BAI Guang-wei;TU Xiu-mei(College of Computer Science and Technology, Nanjing Tech University, Nanjing 211800, China;Nanjing Research and Design Center, Zhongxing Telecommunication Equipment Corporation, Nanjing 210012, China)

机构地区：[1]南京工业大学计算机科学与技术学院,江苏南京211800 [2]中兴通讯股份有限公司南京研发中心,江苏南京210012

出　　处：《计算机工程与设计》2017年第2期449-452,459,共5页Computer Engineering and Design

基　　金：国家自然科学基金项目(60673185);江苏省自然科学基金项目(BK2010548);江苏省未来网络前瞻性研究基金项目(BY2013095-4-09)

摘　　要：为解决当前图像伪造检测算法大多针对单一的伪造形式(复制-粘贴或者拼接),使其通用性弱,检测精度不佳的不足,提出基于Gabor小波变换耦合局部相位量化的图像伪造检测算法。将图像转换为YCbCr色度空间,在Cr通道上进行Gabor小波变换,得到每个尺度和方向的Gabor图像;定义局部相位量化(local phase quantization,LPQ)技术,利用LPQ的相位特征模糊不变性,准确提取每个Gabor图像的LPQ特征值,依据LPQ特征值构建整个图像的特征向量;利用支持向量机进行决策分类,完成图像内容伪造检测。测试结果表明,与当前伪造检测算法相比,所提算法具有更高正确率和检测精度,能够较好适用于不同图像伪造手段的检测。To solve these defects such as weak generality and low detection accuracy induced by single form of forgery(copy paste or mosaic)of current image forgery detection algorithm,an image forgery detection algorithm based on Gabor wavelet transform coupled local phase quantization was proposed The image was converted to YCbCr chrominance space,Gabor wavelet transform was performed in the Cr channel in the chrominance space,and Gabor images on each scale and direction were ob-tained Introducing the local phase quantization(LPQ)technique,the fuzzy invariance of the phase characteristics of LPQ was used to accurately extract the LPQ values of each Gabor image,so the feature vector of the whole image was constructed accor-ding to the characteristic value of LPQ.Support vector machine was used to carry out the decision classification.Experimental results show compared with the current algorithms?this algorithm has higher accuracy,and it can be used to detect different image forgery methods.

关 键 词：图像伪造检测 色度空间 GABOR小波变换 局部相位量化 支持向量机 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]