基于小波纹理分析的隐写分析研究

A novel steganalysis technique based on model is presented. The key element of the method is wavelet coefficients in each subband of wavelet transform are modeled as a Generalized Gaussian distribution ( GGD) with two parameters. These two parameters of each subband coefficients are obtained by the maximum2likehood estimator. It appears that these parameters are a good measure of image features and can be sed to discriminate stego-images from covr images. Neural network , SVM (support vector machine) , regression analysis are adopted to train these parameters to get the inherent characteristic of cover and stego images. Experimental results show that the algorithm is comparable to previously existing techniques. And this method is a general steganalysis method which is applicable for the detection of data hiding and watermarking techniques.