基于图像分割及模糊隶属度的PCA人脸识别

FACE RECOGNITION WITH PCA BASED ON IMAGE SEGMENTATION AND FUZZY MEMBERSHIP

提出一种新的图像分割方法应用于PCA中,将包含人脸特征最为明显的额头、左眼、右眼、鼻子、嘴巴等五部分从图像中分割出来,而舍弃双耳以及脸部其余部分等只包含很少特征的部位。在分类识别中引入模糊隶属方法,提出一个新的隶属度函数并加权融合上述五部分的识别结果。基于ORL人脸库的实验表明,所提出的新分割和隶属度函数结合的方法具有很好的分类效果,提高了识别率和执行效率。

We propose a new image segmentation method to be applied to PCA,which cuts apart five most obvious facial feature parts from the image,including forehead,left eye,right eye,nose and mouth but gives up the parts of two ears and other remained facial areas which containing only a few characteristics. Fuzzy membership is introduced to classification and recognition,and a new membership function is presented,weighted and to fuse it with the recognition results of the above five parts. It is shown by the experiments based on ORL face library that the proposed method combining the new segmentation with membership function has good classification effect and improves the recognition rate and execution efficiency.