摘要
单传感器数码相机一般通过颜色滤波阵列(color filter array, CFA)插值得到彩色图像.利用CFA插值特性检测图像篡改,根据插值图像与理想全色图像的频谱差异分块提取频谱变化和色度失真特征以反映CFA插值特性,计算重插值前后的插值特性变化作为取证特征;使用支持向量机(support vector machine, SVM)分类并根据相邻块间取证特征的不一致来鉴别图像真伪.实验结果表明:该方法可以有效辨别篡改图像,且对JPEG压缩有较好的鲁棒性.
Single-sensor digital cameras generally acquire the missing color components by color filter array(CFA)interpolation.In this work,CFA interpolation characteristics are exploited to identify image forgery.Using the differences in frequency spectrum between interpolated images and ideal full color ones,the interpolation characteristics are described by block spectral change and chrominance artifacts features.The feature difference between test images and their re-interpolated version are computed as forensic features.Finally,support vector machine(SVM)is exploited to classify the authentic and tampered images using the block-wise inconsistency of forensic features.Experimental results verify effectiveness of the proposed method and its robustness against JPEG compression.
作者
苏文煊
方针
SU Wen-xuan;FANG Zhen(School of Communication and Information Engineering,Shanghai University,Shanghai 200444,China)
出处
《应用科学学报》
CAS
CSCD
北大核心
2019年第1期33-40,共8页
Journal of Applied Sciences
基金
国家自然科学基金(No.U1536109)资助
关键词
篡改检测
颜色滤波阵列插值
频谱相关性
支持向量机
forgery detection
color filter array(CFA)interpolation
spectral correlation
support vector machine(SVM)
