Face image classification method, face image classification apparatus and face image classification program

摘要

Each illumination base storage means 1 accumulates in advance illumination bases 12 of each category representative. Each illumination direction estimation unit 2 estimates the illumination direction 13 with respect to the input facial image 11, using the illumination space basis vectors 12, each error norm is minimized. Each error evaluation means 3, respectively calculate the error feature 14, which is a norm error between the input face image and 11 inner product to obtain the inner product of the illumination direction 13 and the corresponding illumination base 12, was determined. Then, based on each feature error 14, category identification means 40, to determine the identity category 15 of the input face image 11.