基于改进支持向量机的隐写分析方法

为了更有效地提高图像隐写分析的速度和正确检测率，提出了一种基于改进的支持向量机的隐写分析方法。采用Fridrich提出的多特征融合提取算法对图像进行特征提取，克服了单一特征不能很好描述图像差别的不足。然后提出了一种将最小二乘法与超球体一类支持向量机（HSOC-SVM）相结合的分类器——最小二乘超球一类支持向量机（LSHS-OCSVM），并与目前广泛使用的FLD和非线性SVM分类器作对比实验。结果表明，方法是一种有效、高速的隐写分析方法。

Steganalysis method based on improved SVM

To enhance the speed and correct examination rate of image steganalysis,this paper provides a new steganalysis method based on the improved SVM.It uses mixture of a few features discussed by Fridrich to extract the features of images,and overcomes the shortcomings that using only one feature can not present image differences well.Then a new classification,Least Square Hyper Sphere One-Class SVM（LSHS-OCSVM） which combines least square programme and the sphere one-class SVM,is provided.Compared with FLD and nonlinear SVM widely used at present,the experiment results prove that it is an effective steganalysis method with high-speed detection.