An efficient feature extraction methodology for blind image steganalysis using contourlet transform and Zernike moments

We propose an effective blind image steganalysis based on contourlet transform and Zernike moments that improves the detection accuracy of universal image steganalysis methods. The proposed method examines randomness in the test image to distinguish between the stego and non-stego images. The suspicious image is decomposed by contourlet transform, and then the absolute Zernike moments of contourlet subbands coefficients of the image and linear prediction error of each contourlet subband are extracted as features for steganalysis. These features are fed to a nonlinear SVM classifier with an RBF kernel to distinguish between cover and stego images. Experimental results show that the proposed features are highly sensitive to the change made by the embedding process to the statistical characteristics of the image. These results also reveal advantage of the proposed method over its counterpart stegan-alyzers, in cases of four popular jpeg steganography techniques. This improvement is mostly due to the greater noise sensitivity achieved using the Zernike moments that yield at least 3.5%–5% higher stego detection accuracy, relating to that of two baseline steganalyzers.