基于Gabor特征的人脸识别算法的对比研究与实现

The Comparative Research and Implementation of based on Gabor Features Face Recognition Algorithms

由于Gabor变换的核函数分布与哺乳动物视觉皮层简单细胞2D感受野剖面非常类似,并具有良好的方向选择性和空间局部性,从而为图像局部区域内多个方向的空间尺度信息和局部性结构特征的获取提供了更有效的方法。为了验证Gabor特征在人脸识别中的有效性和准确性,本文提出了一种采用目前四种传统特征提取的人脸识别方法与基于Gabor特征的人脸识别方法进行对比研究,同时提出利用ROC和CMC两个参量来验证基于Gabor特征人脸识别方法的有效性和准确性。在ORL人脸数据库上取得的实验结果表明,基于Gabor特征的人脸识别方法在同等条件下,得到了更高的人脸识别率,同时具有良好的鲁棒性。

Since the distribution of kernel function of Gabor transform and the 2D receptive field profiles of mammalian simple cells in the primary visual cortex is very similar,and has the direction selectivity and good spatial locality,so the acquisition of spatial scale information of multiple directions and local structure features in the local regions of images provides a more effective method. In order to verify the efficiency and accuracy of based on Gabor features face recognition methods,this paper proposes a comparative research for face recognition by using the methods the present four kind of traditional features extraction and the methods of based on Gabor features extraction,at the same time,using two parameters including ROC and CMC to validate the efficiency and accuracy of based on Gabor features face recognition method. In ORL face database,the experimental results show that the face recognition methods based on Gabor features in the same conditions obtain higher face recognition rate and the robustness.