얼굴 기울기 변동에 강인한 고성능의 얼굴 검출 구조 연구

This paper proposes a study on high performance face detetion architecture for robust in face rotation, which is applicable to robot or security system, etc. To implement and verify the proposed method, we use MCT and extract feature of face and then adopt AdaBoost learning algorithm. The structure is composed of Noise filter, Memory Interface, Image Scaler, Rotation Transformer, MCT Generator, Face Estimator, Position Resizer, Data Grouper and Overlay Processor. Verification shows that number of faces in a frame is maximum 32 faces. For the addition of rotation algorithm to the original, possible detection range of rotated face has been extended and also detection rate, too This achievement is meaningful to overcome illumination intrusion of faces and renders faster and higher detection rates than former implementations.