Real-Time Head Pose Estimation by RGB-D Camera

Abstract

Conventional RGB image-based head pose estimation methods encounter many difficulties due to pose variation and illumination. In this paper, we present a real-time 3D head motion estimation method using both RGB image data and depth data. Head is detected from depth data in a simple way. Based on a rigid-body motion model, we derive the linear depth and optical flow constraint equations respectively. These constraints are combined into a single linear system, from which head motion vector is recovered by minimizing a least-squares. Experimental results have shown that the use of both depth data and RGB data in our method overcomes the shortcomings of single depth or RGB data. In addition, it's still robust when there is only one type of data reliable.