Face Pose Estimation Using Near-Infrared Images

Abstract

Face pose estimation is a research focus in the field of face recognition. In this paper, we present an accurate and fast approach to estimate the pose of face in near-infrared images. First, the approach automatically extracts the following feature points from a normalized face image: two pupils, the nose tip and the center line of mouth. Then we use pupils and nose tip to estimate the pose rotation of yaw, use the pupils, nose tip and the center line of mouth to estimate the pose rotation of pitch. The experiments show that when the facial rotation angle is less than 30 degrees in yaw, 20 degrees in pitch, the estimation error of our approach is less than 5 degrees, and the processing time of each image is about 8ms.