计算机应用 ›› 2013, Vol. 33 ›› Issue (03): 663-666.DOI: 10.3724/SP.J.1087.2013.00663

• 多媒体处理技术 • 上一篇    下一篇

基于四元数多特征并行融合的JPEG隐写检测

何凤英*,钟尚平,杨健   

  1. 福州大学 数学与计算机科学学院,福州 350002
  • 收稿日期:2012-09-17 修回日期:2012-10-22 出版日期:2013-03-01 发布日期:2013-03-01
  • 通讯作者: 何凤英
  • 作者简介:何凤英(1979-),女,福建武平人,讲师,硕士,主要研究方向:信息隐藏、图像处理; 钟尚平(1969-),男,福建武平人,副教授,博士,主要研究方向:网络与信息安全; 杨健(1987-),男,广西南宁人,硕士研究生,主要研究方向:网络与信息安全。
  • 基金资助:

    福建省教育厅科研项目(JB09003)。

JPEG steganalysis based on parallel feature fusion by quaternion

HE Fengying*, ZHONG Shangping, YANG Jian   

  1. College of Mathematics and Computer Science, Fuzhou University, Fuzhou Fujian 350002, China
  • Received:2012-09-17 Revised:2012-10-22 Online:2013-03-01 Published:2013-03-01
  • Contact: Feng-ying HE

摘要: 针对串行特征融合方法易出现“维数灾难”以及并行复矢量特征融合方法只能融合两类特征的弱点,提出一种基于四元数多特征并行融合的JPEG隐写检测方法。方法利用四元数有4个分量能融合4种特征的性质,首先提取4种经典特征,然后用主成分分析(PCA)进行数据降维,去除冗余信息,最后将4种特征组合为四元数矢量,实现多特征的并行融合。实验结果表明,和传统特征融合方法相比,所提方法不仅有效提高了JPEG隐写图像检测率,而且具有较强的鲁棒性。

关键词: 隐写检测, 四元数, 多特征, 并行融合, JPEG

Abstract: Feature fusion has become a heated research topic in steganalysis. In order to solve the problems in the existing feature fusion methods, such as feature dimension disaster in serial feature fusion method and only two kinds of features can be integrated in parallel feature fusion method, this paper proposed a JPEG steganalysis approach based on parallel feature fusion by quaternion. The method firstly selected the four fusion features, and then isolated redundant components by the Principal Component Analysis (PCA), finally combined four features as quaternion vector to achieve multi-feature parallel fusion using the four components of the quaternion. The experimental results show that, compared with the traditional feature fusion method, the proposed method effectively improves the detection rate of steganalysis in JPEG images, and has better robustness.

Key words: steganalysis, quaternion, multi-feature, parallel fusion, JPEG

中图分类号: