基于统计特征的人脸识别研究

奇异值特征向量是用于图像识别的有效代数特征，但直接用奇异值特征向量做匹配进行人脸识别，识别率极低。通过对人脸图像奇异值向量和其对应的左右正交特征矩阵分析，发现图像的奇异值向量与图像的灰度范围具有相关性，即最大奇异值反映了图像灰度范围的位置，其他奇异值反映了灰度范围的宽度，而且与图像奇异值向量对应的左右正交特征矩阵能够表现图像轮廓的结构信息。基此，提出基于奇异值分解（singular value distribution，SVD）的基空间人脸识别算法,并通过ORL和ORL-IC数据库进行仿真，实验结果分析证明了图像的左右正交特征矩阵能够表现图像轮廓的结构信息。

Research of face recognition based on statistical image feature

According to the analysis of the singular value vector of face image and left-right orthogonal characteristic matrix which corresponds to the singular value vector, the singular vectors related to image grayscale was found in the paper as follows: The maximum of the singular values presented the position of image grayscale, and other singular values showed the width of Grayscale. Moreover, it was also discovered that the left-right orthogonal characteristic matrix which corresponded to image singular value vectors could present structure information of picture contour. Finally, basis space face recognition algorithm based on SVD(singular value distribution) was proposed in order to make further certification that image left right orthogonal characteristic matrix could present structure information of picture contour, carrying on emulation in the ORL and ORL-IC database. It is concluded that the left-right orthogonal characteristic matrix can present structure information of picture contour.