Classification between PS and Stego Images Based on Noise Model

Owing to the popular usage of Photoshop, PS images which are processed by Photoshop or other similar software nowadays emerge increasingly. PS images may be misfortunes to steganalysis. The similarities and differences between PS and stego images are analyzed in this paper, and a method is proposed to classify the natural, PS and stego images. The first-scale diagonal subband obtained by wavelet transform is decomposed, and then the diagonal decomposed subband is decomposed again, the statistical moments of characteristic function (CF) of the image and its wavelet subbands are selected as features. Artificial neural network is utilized as the classifier. Extensive experimental results showed that the proposed method can classify natural images, common PS images and typical stego images reliable. Further more, the sharpening and contrast enhancement images can also be classified.