中文摘要：

基于小波域高频系数的广义高斯分布（GGD）模型，提出了一种新的特征提取优化算法。通过分析研究小波系数概率密度函数的时频特性，选择在频域中提取高阶统计矩。为了更好地区分原始图像和隐秘图像高阶矩的差别，对高阶统计矩的权重函数进行改进。除此之外，对高阶统计矩阶数的确定，以及子带层次的选择也作了进一步的研究，构建了提取最优特征矩的提取算法。基于Matlab7．0平台进行仿真实验，结果证明：该算法的综合性能明显优于同类特征提取算法。

英文摘要：

A new optimized algorithm for feature extraction based on GGD model of wavelet coefficients is brought forward. Analyzes time - frequency domain character of the PDF of wavelet coefficients and chooses to extract the higher - order.statistical moments from the frequency domain. For discriminating the differences between cover image and stego image, improve the weigh function and it makes the effect of extraction better. Besides that,analyzes how to select the order of higher-order moments and hierarchy of wavelet subband. Finally,make emulational experiment on the Matlab7.0 platform, experiment results prove that the extraction algorithm has better performances than the congeneric algorithms.