Two New Methods Of Face Recognition

We present which we call Image Matching Distance(IMMD).This distance considers the relationship between the every point of image and the specific area of corresponding image, finds matching point in this special area, lets the image of the gray level and its location introduced into the similarity measure of image. It makes IMMD have a good robustness for the changes of face posture, angle, and the expression. Embedding IMMD in kernel Fisher discriminant analysis (KFDA) for face recognition. The experimental results show that this method is superior to the same type method which embedded Traditional Euclidean Distance and Image Euclidean Distance. According to the idea of kernel principal component analysis(KPCA) and membership degree, we also present a new method of image feature extraction, which is Class Information Incorporated Fuzzy Kernel Principal Component Analysis Method(CIFKPCA).CIFKPCA defines a appropriate membership degree to describe the sample distribution, and uses CIFKPCA extract the features can include gray level information and distribution information. The experimental results on two Face Database show that CIFKPCA is superior to KPCA method and even superior to KFDA method.