应用五株采样提升算法的抗盲检测图像隐写算法  

Blind Detection Resistant Steganographic Algorithm for Images Based on Quincunx Sampling Lifting Scheme

作　　者：陶然[1] 张涛[1] 平西建[1] 

机构地区：[1]解放军信息工程大学信息工程学院,郑州450002

出　　处：《数据采集与处理》2012年第2期179-188,共10页Journal of Data Acquisition and Processing

基　　金：国家自然科学基金(60903221)资助项目

摘　　要：在分析图像盲检测算法原理的基础上,提出了一种在形态小波高频系数上进行消息嵌入的抗盲检测隐写算法。该算法利用五株采样提升实现图像的小波变换,在大于一定门限的小波高频系数中嵌入消息,并通过建立的嵌入信息表来修正嵌入规则以保持小波系数直方图近似不变,在门限处引入直方图调整策略以减小系数直方图在门限处的变化。由于通用盲检测算法大多基于概率密度函数的变化实现图像隐写的检测,因此本文算法可以获得对通用盲检测算法的抵抗能力。实验结果表明,本文算法在抵抗小波高阶统计量分析、直方图特征函数质心等盲检测算法能力方面,优于LSB匹配、像素值差分等隐写算法。Based on analysis of the principles of blind detection techniques, a data hiding algorithm by modifying morphological wavelet high frequency coefficients is proposed. The quincunx sampling lifting scheme is used for image decomposition. Then the subband coefficients above a certain threshold are chosen for data embedding, and the embedding information table is built for amending algorithm. Moreover, the histogram adjustment strategy is introduced at the location of threshold coefficients to preserve the histogram of wavelet coefficients. Since most blind detection algorithms select classifying features according to the differences of statistical distributions between cover and stego images, the proposed method can resist the attack of blind detection techniques. Experimental results show that the proposed method outperforms previous steganographic methods, such as least significant bit （LSB） matching and pixel-value differencing in the capability of resisting current typical universal blind detecting methods.

关 键 词：数字隐写 通用盲检测 形态小波 五株采样 提升算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]