文档介绍：Face Detection Based on Kernel Fisher Discriminant Analysis
Yuanjian Feng, Pengfei Shi
Institute of Image Processing and Pattern Recognition
Shanghai Jiaotong University
Shanghai 200030, China
Abstract 2. Fisher’s Linear Discriminant and the
Kernel-based Extension
This paper presents a face detection method based on . Fisher’s Linear Discriminant
Kernel Fisher Discriminant analysis (KFD). Kernel based
methods have been extensively investigated both in theo- Fisher’s Linear Discriminant Analysis (LDA) is a classi-
ries and applications, such as SVM and Kernel PCA. Using cal classification method. Given samples from two classes,
½ ¾
½ ¾
½ ¾
Ü
Ü
Ü Ü
Ü Ü
¾
½ and ,
½ ½
¾
the kernel trick, Linear Fisher Discriminant can be ex- ¾
Æ Æ
½ ¾
tended to non-linear case. Since the distribution of face LDA finds an optimal direction Û along which the dif-
patterns is plex and highly nonlinear, using non- ference between the two classes is maximized while
linear classification tools can hopefully tackle the prob- the variance of each class is minimized. Û can also be
lem of face detection. We explore the application of KFD seen as a linear transform which maps the input space
Ñ
in the task of frontal face detection. The experimental re- onto an one-dimension transformed space[2]. Let Ð
È
Æ
Ð
½
Ð