Computer Engineering and Applications ›› 2013, Vol. 49 ›› Issue (24): 70-73.

Previous Articles     Next Articles

Blind extraction algorithm of embedded messages based on CSR-ICA

FENG Fan1,2, WANG Jianhua1, ZHANG Zhengbao2, WANG Huiping1, QI Hongjun1, ZHANG Hailong1   

  1. 1.Department of Basic Courses, Bethune Military Medical College, Shijiazhuang 050081, China
    2.Department of Computer Engineering, Ordnance Engineering College, Shijiazhuang 050003, China
  • Online:2013-12-15 Published:2013-12-11

基于CSR-ICA模型的隐写信息盲提取算法

冯  帆1,2,王建华1,张政保2,王惠萍1,戚红军1,张海龙1   

  1. 1.白求恩军医学院 基础部,石家庄 050081
    2.军械工程学院 计算机工程系,石家庄 050003

Abstract: Aiming at steganography adding model, a blind extraction algorithm of embedded messages is proposed based on CSR-ICA. It attains estimated signal of original cover signals using linear combination just one digital image, which consists with constrain of ICA model. Then signals with optimal normalization kurtosis are selected and inputted to ICA model by optimization algorithm. In order to eliminate effect of outlier, normalization kurtosis is taken as objective function of separation algorithm on precision. The algorithm has higher comprehensive performance and conquers the localization of proposed algorithm by Chandramouli. Average extraction correct rate is 90%. The results of simulation experiments prove its validity further.

Key words: Contourlet Sparse Representation(CSR), Independent Component Analysis(ICA) model, steganography, blind extraction

摘要: 针对加性隐写模型,提出一种基于CSR-ICA的隐写信息盲提取算法。算法仅需一幅隐写图像,在满足ICA模型线性约束条件下得到载体信号的估计信号,通过Contourlet稀疏性表示(CSR)对模型输入信号进行前置处理,优化选取归一化峭度性较大的信号作为模型输入信号,将归一化峭度作为分离算法学习的目标函数,避免异常值给分离算法带来的误差。算法具有较好的综合性能,并且克服了Chandramouli算法的局限性,提取正确率平均为90%。仿真实验结果给出了算法的有效性验证。

关键词: Contourlet稀疏性表示(CSR), 独立成分分析(ICA)模型, 隐写, 盲提取