基于图像纹理复杂度和奇异值分解的重采样检测

Detection of Resampling Based on Texture Complexity and Singular Value Decomposition

摘要: 针对图像篡改时可能会经历重采样操作,而重采样过程中的插值步骤会对重采样图像像素引入一定的线性相关性的问题,提出一种重采样检测算法.采用奇异值分解度量像素间的线性相关性,针对纹理复杂程度不同的子像素块分析插值处理对其线性相关性的影响;以零奇异值个数和奇异值均值作为分类特征,结合SVM进行重采样检测.实验结果表明,该算法能够实现对重采样图像和原始图像的准确分类.

Abstract: When images are tampered,they may undergo resampling manipulation.A resampling detection algorithm is proposed,which is designed to detect the changes brought into the linear dependencies of the image pixels introduced by the interpolation process of resampling.The specific statistical changes with different texture complexity are analyzed by singular value decomposition,the number and mean value of small singular values are used as distinguishing features,and SVM is used to classify.The performance of resampling detection shown in experimental results indicates the validity of the algorithm.