Fast Robust Face Detection under a Skin Color Model with Geometry Constraints

Face detection is an important first step towards solving plethora of other computer vision problems. A fast robust face detection algorithm under a skin color model with geometry constraints is proposed in this paper. An ellipse color model in the YCbCr color space is first used to detect the skin regions. A light compensation algorithm is utilized to the original image, followed by a box model to detect hair regions. The detected candidate skin and hair regions are then combined with geometry constraints to detect faces in the image. Experiments show that the proposed method is effective and efficient.