针对互联网上大量的隐写软件利用数字图像隐藏秘密消息,文中提出了基于净图描述的通用隐写分析技术.这种隐写分析技术能有效地检测出图像是否含秘密信息,通过提取两种新颖特征,即纹理特征和虚特征值分解(IED)特征,对净图进行完整的定量描述,又经过降维处理,解决了维数灾难问题,提高了检测速度;并且设计了与特征相匹配的柔性化超椭球体一类分类器,提高了检测性能.
On Internet,There are a large number of steganographic programs to hide message in digital image.This paper proposes a novel general steganalysis technique based on cover image describing.It can detect effectively whether or not secret messages are in digital image by this technique.Cover image is quantitatively described fully by texture feature and imaginary eigenvalue decomposition(IED)feature.It can increase detection speed by a reducing dimension processor,and the curse of dimensionality problem can be solved.One classifier based on a flexile super-ellipse-sphere is designed,and detection performance is improved.