Detection of faces under shadows and lighting variations

Abstract

Successful face detection and facial feature extraction is crucial for for a variety of applications, including speech recognition and fatigue monitoring. Detecting and tracking faces in a moving automobile is challenging because of a variety of reasons. Among these reasons are changing poses, extreme lighting changes and shadowing. In this paper, we investigate two approaches for shadow compensation in such images. Together, these techniques offer a good improvement in the face detection accuracy on a data set captured in a moving automobile.