提出了一种基于小波包分解和小波系数相关性的通用型图像隐写分析方法。对差分处理后的图像进行小波包分解,提取图像及其小波子带的高阶统计量作为特征。利用图像小波系数在尺度方向和空间方向的相关性,使用马尔可夫模型挖掘小波系数层内和层间相关性,提取转移概率矩阵作为特征。针对H4PGP、F5和OutGuess隐写算法的实验表明,方法对上述三种隐写算法具有较好的检测效果。
In the paper, a new image steganalysis technique using wavelet packet decomposition and wavelet coeffi- cient dependencies is proposed. It decomposes the preprocessing image using wavelet packet decomposition and ex- tracts the higher-order statistics of the image and its wavelet packet sub bands as features. It exploits the scale and orientation dependencies between wavelet coefficients and extracts intrascale and interscale features using Markov matrices. The experiments on H4PGP, F5 and OutGuess steganography show the method can detect stego images reliably.