Study on dependencies between wavelet coefficients for image steganalysis and its application
图像隐写分析中小波系数相关性的研究与应用*

The paper proposed a new blind image steganalysis scheme for spatial domain image steganography. It exploited mutual information to analyse the change on the scale and orientation dependencies between wavelet coefficients which was caused by the embedding of a message. Applied Markov model to model the dependencies between wavelet cofficients so as to extract intrascale and interscale transition probability matrices. The experiments on LSB matching and stochastic modulation steganography show the method can detect uncompressed stego images reliably. For low embedding rate stego images, the detection accuracy of this method exceeds those of its closest competitors by at least 8% and up to 14%.