시선정보 분석을 통한 관심영역 기반의 얼굴검출 성능 향상

In this paper, we propose a new face detection model which hybrids the biologically inspired selective attention models with the Adaboost based face detection method. We have developed a glass type eye tracker using two cameras in which one is for monitoring of outside environment and the other is for detecting human eyeball. The proposed method can select a candidate area of human faces based on selective attention model and calibration process with eyeball detection, and zoom-in the selected area for detecting human faces. Experimental results show that the proposed method efficiently detects the different sizes of human faces.