基于局部特征分析的人脸识别方法
Face Recognition Based on Local Feature Analysis
-
摘要: 在传统的弹性图匹配基础上,提出一种基于局部特征分析的人脸识别算法.该方法利用人脸的先验结构和人脸图像的灰度分布知识,首先粗略地找出人脸图像的特征点,然后利用人脸弹性图对特征点的位置进行调整,最后在各个特征点处计算Gabor变化的系数,人脸相应被表示为特征点处的Gabor系数集合.对提取的特征向量用几种不同的度量距离来进行分类,并给出测试结果.实验表明,该方法优于传统的Eigenface方法,特别适用于训练图像样本较少的情况.相对于传统的弹性图匹配方法,该方法由于人脸特征点预先被估算出,而不是在整个图像上搜索,所以大大减少了计算量.Abstract: Aface recognition method based on feature analysis is presented here. The local features are first located by the face structure knowledge and gray level distribution information, rather than searching the whole image as it does in Elastic Bunch Graph Matching. The coarse positions of the features are located, and then we use a data structure named Face Bunch Graph to adjust the feature positions. After the accurate positions of the features are located, the face is represented by the Gaborjets, which calculated on these local feature positions. Several distance metrics are tested and the results are given. In our experiment, this method works better than the Eigenface, and consumes less ti me than the traditional Elastic Bunch Graph method.
下载: