METHOD AND SYSTEM FOR DETECTING FACE USING PATTERN SORTER LEARNED THROUGH FACE/SIMILAR FACE IMAGE

摘要

PURPOSE: A method and system for detecting a face using a pattern sorter learned through face image is provided to improve reliability for detecting a face by learning a face and a similar face pattern in a pattern sorter in advance. CONSTITUTION: A face DB(100) includes face images. A similar face DB(105) includes an image of a pattern similar to a face. A base vector extracting unit(110) receives a plurality of images from the face DB(100) and the similar face DB(105) and extracts a plurality of base vectors which express images. A feature extracting unit(120) projects a face image and a similar face image stored in a plurality of base vectors and detects a feature vector of an image and a remaining error. A clustering unit(130a) classifies feature vectors being supplied from the feature extracting unit(120) into a predetermined sub class. A feature class database(140) stores feature vectors according to each sub class. An SVM(Support Vector Machine) learning unit(150) learns SVMs according to each sub class.