Head pose estimation based on image abstraction for multiclass classification

Abstract

We address the problem of head pose estimation from a facial RGB image as a multiclass classification problem. Head pose estimation continues to be a challenge for computer vision systems due to extraneous characteristics and factors that do not contain pose information and affect changing pixel values in a facial image. To achieve robustness against variations in identity, illumination condition, and facial expression, we propose an image abstraction method that can reduce unnecessary information and emphasize important information for facial pose classification. Experiments are conducted to verify that our head pose estimation algorithm is robust against variations in the input images. © 2013 Springer Science+Business Media Dordrecht.