基于像素相关性的数字图像隐写分析

摘要

According to the adjacent pixels of natural images have certain correlation, proposed a steganalysis method which based on correlation measurement of image pixels. Firstly, used Hilbert space filling curve scanning pixels of the image and pixels of each planes, constructed one-dimensional sequence. Then correlated the correlation coefficient and the ratio of the difference between the adjacent pixels to measure the correlation. Extracted the features to input the classifier. The experiments prove the algorithm can effectively detect JSteg, F5 and OutGuess algorithms.%根据自然图像相邻像素之间具有一定相关性的特点,提出一种基于图像像素相关性度量的隐写分析方法.首先利用Hilbert空间填充曲线对图像各位平面像素以及图像像素进行扫描,将其构造成一维序列,然后分别采用自相关系数和像素前后差值比对图像相邻像素间的相关性进行度量,提取特征输入到分类器中.实验表明,该算法能够实现对JSteg、F5、OutGuess三种隐写算法的有效检测.