基于场论聚类的集成分类器隐写分析算法

Steganalysis based on Field Theory and Ensemble Classifiers

针对目前隐写分析中特征空间有待提高的问题,提出一种基于场论聚类的隐写分析算法。算法提取共生矩阵特征、差分特征等多维特征用以描述图像,再利用场论进行聚类,从中选择代表性特征用以训练个体分类器,最后通个加权融合法完成对个体分类器进行集成。实验表明,该算法能够有效提高隐写分析的检测率。

To improve the accuracy in steganalysis, this paper proposed a steganalysis based on field theory and ensemble classifiers.It adopts algorithms of clusterng based on field theory to train single classifiers with excellent diversity. Then it put forward improved feature weighted ensemble classifiers as a new classification to enhance the examination rate. The superiority of this algorithm in classification performance e is established extensively through experiment.