Journal of Computer Applications ›› 2014, Vol. 34 ›› Issue (1): 113-118.DOI: 10.11772/j.issn.1001-9081.2014.01.0113

• Network and distributed techno • Previous Articles     Next Articles

Universal blind detection based on multiple classifiers fusion and image complexity

WANBaoji1,ZHANG Tao1   

  • Received:2013-05-14 Revised:2013-07-17 Online:2014-01-01 Published:2014-02-14
  • Contact: WANBaoji



  1. 1. 信息工程大学 信息系统工程学院,郑州 450002
  • 通讯作者: 万宝吉
  • 作者简介:万宝吉(1986-),男,甘肃庄浪人,硕士研究生,主要研究方向:信息隐藏、信息融合;张涛(1977-),男,湖北天门人,副教授,博士,主要研究方向:信息隐藏、图像处理、模式识别。
  • 基金资助:


Abstract: The current blind detection techniques do not consider how the contents of different images influence the steganalysis performance. In this paper, a new approach based on image content and classifier fusion was proposed. In the training phase of the proposed method, the input images were first divided into several classes according to the image complexity, the training process was specialized and then the fuzzy measure was calculated for each class. In the testing phase, the class of image was first obtained, and various classified results were acquired by classifiers and then a fuzzy integral was used to fuse different classes in the decision making process. The experimental results on several sets of images demonstrate that the proposed steganalyzer significantly enhances the detection accuracy of prior art.

Key words: steganalysis, universal blind detection, image complexity, fuzzy integral, classifier fusion

摘要: 现有通用盲检测方法大多没有考虑图像内容对隐写分析性能的影响,对此提出一种利用图像内容复杂度进行预分类和多分类器融合的隐写分析方法。在训练阶段,首先根据图像复杂度把图像分为若干类,然后针对每一类别训练分类器,并计算其模糊测度。在测试阶段,先判断待测图像的类别,然后将其送入到已训练好的各个分类器中,得到多个局部决策值,之后对其进行模糊积分融合得到最终的检测结果。实验结果表明,所提方法提升了通用盲检测算法在混合图像库中的检测性能。

关键词: 隐写分析, 通用盲检测, 图像复杂度, 模糊积分, 分类器融合

CLC Number: