基于2DGabor的人脸识别改进算法

Improved face recognition algorithm based on 2DGabor

为解决已有PCA人脸识别算法对人脸偏转和光线变化不具有鲁棒性以及易丢失人脸结构信息的不足,提出一种基于2DGabor的改进人脸识别算法。利用2DGabor提取人脸初始特征,生成Gabor图像;利用旋转不变一致LBP算子对Gabor图像进行特征提取,分块统计特征向量;通过2DPCA算法降低特征维度,建立特征脸,输入分类器,进行匹配识别。在VS2013+OpenCV下,选用ORL、YaleA、YaleB和CMU-PIE人脸库为实验数据集,实验结果表明,与已有算法相比,在面部偏转和光线变化情况下,该算法缩短了匹配时间,提高了识别率。

To solve some deficiencies of existing face recognition algorithms on PCA, such as their non-robust performance in the face rotation and lighting change, and lost facial structures, an improved face recognition algorithm based on 2DGabor was proposed. 2DGabor was used to extract initial facial features, and Gabor images were generated. The features of Gabor images were extracted using the rotation invariant uniform LBP operator, and eigenvectors were segmented into blocks. The 2DPCA algorithm was used to reduce the facial feature dimension, to establish the feature face, and to input the classifier for matching face recognition. Experiments were carried out on ORL, YaleA, YaleB and CMU-PIE database using OpenCV in VS2013. Experimental results show that the proposed algorithm can improve the recognition rate and shorten the matching time facing the face rotation and lighting change compared with existing face recognition algorithms.