基于SVM的多特征图像隐秘分析  

Image steganalysis based on multi-feature extraction using support vector machines

作　　者：李钢[1] 张国良[1] 张仁斌[1] 

机构地区：[1]合肥工业大学计算机与信息学院,安徽合肥230009

出　　处：《合肥工业大学学报（自然科学版）》2007年第4期424-427,共4页Journal of Hefei University of Technology：Natural Science

摘　　要：文章通过分析研究隐秘图像和正常图像小波子带系数高频部分的统计特征,从纹理统计矩、DCT系数直方图矩和上下文块之间的相关性方面来提取特征,组成特征向量,并采用SVM方法进行分类,实现了一种从不同类型、不同角度提取多个特征的图像隐秘分析算法,解决了现有通用性隐秘分析算法特征提取不足的问题。实验结果表明,这是一种有效的、高精度的及通用性的检测方法。The statistical features of high frequency stego-images are analyzed,and the feature vectors wavelet sub-band coefficients of natural images and are extracted from the texture statistical moment, the DCT coefficients＇ histogram moment and correlativity between blocks, and then the stego-images are classified by the SVM method. The proposed steganalysis algorithm extracts multiple features from different types and perspectives,thus solving the insufficiency of feature extraction in current universal steganalysis methods. Experiment results show that the new algorithm is an effective universal detection method with high accuracy.

关 键 词：隐秘分析 多特征提取 支持向量机 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]