METHOD OF DETECTING POSITION OF EYES IN VIDEO DATA OF VARIOUS BACKGROUNDS

摘要

PURPOSE: A method of detecting the position of eyes in video data of various backgrounds is provided to receive various face images using templates matching to independently detect the position of eyes without being restricted by race, disguise, photographing angle, rotation and so on. CONSTITUTION: A moving region is extracted from video data, inputted in real time, using a difference image regardless of a variation in a background image. A motion using the moving region as a detection candidate region is detected. A skin color is extracted from the extracted region using HSI color. Templates matching is carried out while the detection candidate region is being scanned using a predetermined average image, to detect a face. The position of the eyes of the face is detected in order to normalize the face image through accurate size correction and rotation correction.