Three-Dimensional Head Pose Estimation Using a Stereo Camera Arrangement

Abstract

Head-pose estimation is a crucial component for analysing human behaviour through various 2D and 3D applications. However, the usage of the strategies based on 2D technologies is not very effective, as the sources of data are limited. In contrast, the usage of the strategies based on 3D technologies is a promising area. The 3D HPE methods are also imperfect, especially when they are applied in situations of inconsistent illumination or occlusion. An analysis of related works helps to establish an appropriate framework for the current research, and this paper extends previous work by creating an algorithm that further improves the framework. The method relies upon a stereo camera arrangement and utilises a method to detect and track key landmark points of the human face to evaluate improvement in 3D-head-pose estimation.