A statistical steganalysis algorithm is proposed based on the correlation analysis and magnitude evaluation of wavelet coefficients.Firstly
the local regularity of wavelet coefficients correlation is quantitatively analyzed.Secondly
wavelet coefficient magnitudes of the selected regular local areas are evaluated adaptively
using both the inter-scale and intra-scale statistical correlations of the natural image wavelet coefficients.Finally
stego-images are distinguished from cover-images by means of analyzing the statistical property of the prediction errors.The proposed method proved to be highly effective for the wavelet domain steganography.