顾翔元, 郭继昌, 田煜衡, 李重仪
AuthorsHTML:顾翔元, 郭继昌, 田煜衡, 李重仪
AuthorsListE:Gu Xiangyuan, Guo Jichang, Tian Yuheng, Li Chongyi
AuthorsHTMLE:Gu Xiangyuan, Guo Jichang, Tian Yuheng, Li Chongyi
Unit:天津大学电气自动化与信息工程学院，天津 300072
Unit_EngLish:School of Electrical and Information Engineering, Tianjin University, Tianjin 300072, China
Abstract_Chinese:隐写检测特征维数的增加, 不仅增加了分类器训练时间和预测时间, 甚至还会造成“维数灾难”．因此, 为达到特征降维的目的, 对空域隐写检测特征选择进行研究, 提出了一种基于条件互信息的特征选择算法．该算法首先选取一个与类标签具有最大互信息的特征, 接着选取与此特征和类标签具有最大条件互信息的一个特征; 然后通过前向寻找方式, 从未选择特征子集中循环选取与刚选取特征和类标签具有最大条件互信息的特征, 一直到选出规定数目的特征后结束循环．实验结果表明, 与其他算法相比, 所提算法取得了较好的特征选择效果．
Abstract_English:The increase in dimensions of steganalytic features not only increases the time of classifier training and classifier prediction，but also can cause “the curse of dimensionality”. Therefore，in order to decrease the dimensions of steganalytic features，spatial-domain steganalytic feature selection was investigated and a feature selection algorithm based on conditional mutual information was proposed. The proposed algorithm first selects the feature which has the maximum condional mutual information with class label. Then，it selects the feature which has maximum conditional mutual information with the class label and the previous selected feature. Following that，from the candidate feature set，it loops to select the feature which has maximum conditional mutual information with the class label and the previous selected feature in a forward search way. The loop ends when the specified number of features is selected. The experimental results show that the proposed algorithm performs better than other algorithms.
Keyword_Chinese:特征选择; 隐写分析; 空域隐写检测特征; 条件互信息
Keywords_English:feature selection; steganalysis; spatial-domain steganalytic feature; conditional mutual information

PDF全文下载地址:http://xbzrb.tju.edu.cn/#/digest?ArticleID=5912