融合局部gabor相位特征和全局本征脸的人脸识别算法

Face Recognition Based on Fusion of Local Gabor Phase Characteristic and Global Intrinsicfaces

提出一种融合局部gabor相位特征和全局本征脸特征进行人脸识别的方法.该方法采用多个分类器的集成,首先利用gabor滤波良好的空间位置与方向选择特性,用gabor滤波器对图像进行滤波,采用局部XOR算子提取滤波图像的局部gabor相位特征,通过Fisher判别式对每个频率和方向下的相位特征进行降维,融合各个频率和方向下的分类概率,得出局部特征分类信息;然后利用本征判别式方法,得出人脸图像的全局分类信息;最后融合局部和全局分类信息进行识别.通过在三个人脸库中的实验结果显示,本文提出的方法具有很好的识别性能.

A new face recognition method, which is based on fusion of local gabor phase characteristic and global intrinsicfaces is de- veloped in this paper. There are three steps in our proposed method. Firstly, According to the good spatial position and orientation of gabor filter, a gabor filter is applied to filter face images. Local gabor phase characteristics are extracted and fisher linear discriminant analysis is used to project these charactersitics of each spatial position and orientation into low dimensional space. Local classifying scores are developed by fusion of scores in all the spatial and orientations. Secondly, Intrinsic Discriminant Analysis （ IDA ） which tries to best classify different face images by maximizing the individuality difference, while minimizing the intrapersonal difference is applied to get the global classifying scores. Finally, the local scores and global scores are fused to get the classifying result. Experi- mental results on three dataset show that our method can have good recognition performance.