Face Detection at a Distance Using Saliency Maps

Abstract

Face detection has been one of the most studied topics in the biometrics literature. A critical factor in the face detection process is the complexity of the setup used to acquire the images. The face detection might be considered a solved problem in easy setups but for outdoor setups with unconstrained illumination and background conditions the problem is still hard. In this paper we propose using saliency maps as a clue for false positive rejection in the applications that require face detection at a distance in complex unconstrained outdoor backgrounds. The saliency maps will be used in a robust score fusion approach combined with other aspects like facial features and skin color. The advantage in using saliency maps is that it tackles the problem from another perspective that is not affected by pose, illumination and expression problems. Experimental results show that the proposed method is promising.