Detecting faces accurately under natural lighting conditions remains a challenging task for advancing the performance of face recognition systems. Most face detection algorithms are associated with visible spectrum imagery which compromises their efficiency when image acquisition is subject to variable illumination. In this study, we propose a simple yet powerful solution that uses multiple cues to detect and locate faces accurately in near infrared (NIR) imaging systems. Specific objectives include (I) extract optimally the illumination and reflectance components of the original image using an intrinsic images decomposition technique, (2) modify the segmentation process by using an energy minimization procedure in an adaptive fashion, and (3) propose an enhanced face detection process on the basis of pupil localization. The evaluation of our face detector on image sequences of 45 different subjects from the CBSR NIR face dataset demonstrated a highly competitive accuracy with less than 5% error rate.