Steganalysis of Color Images Based on Noise Model and Channels Integration

Steganalysis of color images weaken or ignore the correlation among different color channels by only using single signal channel. The noise model of stego color images is analyzed, and a general steganalysis algorithm based on noise model and color channels integration is proposed. Firstly, the wavelet decomposition of the image is made, and a filtering operation is applied to obtain the wavelet subbands of noise image. Secondly, noise gradient orientation sequences between any two noise channels and noise gradient sum sequence are extracted from the noise wavelet subbands. Thirdly, color gradient orientation sequences between any two channels and color gradient sum sequence are extracted from the color image. Finally, the HilbertHuang transform based vibration features of these sequences are integrated as eigenvectors, and SVM is used to detect images. The experimental results show that the proposed technique realizes the reliable steganalysis of color images with higher correct rate and lower false positive rate, compared with traditional color image steganalysis algorithms.