METHOD FOR RECOGNIZING FACE USING TWO-DIMENSIONAL CANONICAL CORRELATION ANALYSIS

摘要

A method for recognizing a face is provided to calculate a correlation variable vector with front-side face data, corresponding to illumination or pause learning data, through the two-dimensional canonical correlation analysis. A method for recognizing a face includes the following steps of: calculating a correlation variable vector for maximizing the correlation through the two-dimensional canonical correlation analysis, wherein the correlation is the relation between face images or user face pose images under various illumination states and stored front-side face image data; normalizing an inputted image to fit the position of eyes and the image size so as to be identical to learning data; and performing the recognition of a front-side face stored in a DB(510) by using a value obtained by the dot-product of the correlation variable vector and a pre-processed input image.