摘要
环形对称Gabor变换不但继承了传统Gabor小波的一般特性,而且冗余性小,具有严格旋转不变性。首先将图像映射到环形对称Gabor变换(CSGT)域,然后用AdaBoost算法挑选出最能表征人脸且有较好分类效果的CSGT特征,同时减少特征向量的维数,最后使用加权PCA算法对特征进行分类。在ORL人脸库上进行实验,结果表明,与传统人脸识别算法相比,该算法对光照、姿态等影响因素具有更好的鲁棒性。
Circularly symmetric Gabor transform （CSGT） not only inherits the general characteristics of traditional Gabor wavelet, and also has the advantages of low redundancy and has strict rotation invariance. In this paper, we propose a novel face recognition approach based on CSGT and AdaBoost algorithm. The face images are mapped onto the CSGT domain first; and then the CSGT features are extracted using the AdaBoost algorithm that can best represent the faces and has a better discriminant performance; meanwhile the dimensions number of eigenvector are reduced as well. Finally the features are classified employing the weighted PCA algorithm. Experimental results on the ORL face database show that the proposed approach is more robust to the variation on illumination, pose, etc. compared with existing approaches.
出处
《计算机应用与软件》
CSCD
北大核心
2012年第7期6-8,13,共4页
Computer Applications and Software
基金
国家自然科学基金项目(60872119)
