基于局部保持映射的图像隐密检测算法

提出了一种新的基于局部保持映射（Locality Preserving Projections，LPP）降维的图像隐密检测方案。为降低图像特征向量的维数，同时保持其内在低维结构，方便构造更有效的分类器，在经过小波变换形成图像特征后，利用LPP算法得到图像特征集的低维流形，实现对图像高维特征的降维。进而使用支持向量机（SVM）算法将降维后的特征映射到分类特征空间，实现对正常图像和隐密图像分类。实验结果表明，与不采用降维算法的检测方案相比，提出的方案能够显著地提高检测的准确率。

Image steganography detection algorithm based on locality pre-serving projections

A new image steganography detection algorithm based on locality preserving projections has been intreduced,In order to reduce the dimensionalities of the image feature set,keep its low dimensional structure and make the classifier more effective, after forming the image features set via wavelet transform,the algorithm projects the set to a low dimensional one and gets the inner manifold by using the LPP algorithm.Support vector machine classifier is used to classify the image set.It is testified that in contrast with algorithm without dimensionality reduction,this algorithm can improve the performance signature signally.