基于改进局部纹理特征的隐写检测算法

图像隐写是一种将信息隐藏于数字图像中的技术,而隐写检测算法试图分辨出藏有信息的图像。文中分析了现有隐写检测算法的优势,结合HUGO隐写算法的特点,提出了一种改进的基于局部纹理特征的隐写检测算法。该算法由LOCP和LPQ这两种局部纹理特征组成。由于提取到的特征维度很大,因此选用了Ensemble分类器进行训练与检测。在HUGOBOSS 1.0图像库上的实验显示,提出的隐写检测算法比原有算法更准确地区分出原始图像和隐写图像,并获得了83.65%的检测准确率。

A Steganalyis Scheme based on Improved Local Texture Features

Steganography is a technique that hides information into digital images, while steganalysis tries to distinguish whether there are data in images. This paper analyzes advantages of current steganalysis schemes, and takes the particularity of HUGO steganography algorithm into account, proposes an improved approach to steganalysis with local texture features. It is mainly composed by two kinds of local texture features： LOCP and LPQ. As a result of high dimension features, Ensemble classifier is used for training and testing. The experiment on HUGOBOSS 1.0 image database indicates that, the proposed scheme performes better in distinguishing original images and cover images, and could achieve 83.65% in detection accuracy.