Human face image analysis

Abstract

Human face detection is the first procedure in applications such as drowsiness detection for drivers, video surveillance, perceptual quality in 3G mobile video communications, detections in digital cameras, face recognition, image compression, image database management and computer animation. We propose real time face detection in real time video captured by a web cam, in complex backgrounds and varying lighting conditions based on the YCrCb color model, which is widely used in video compression standards. Our method first finding the face candidates using boosted classifier (statistical model) cascades of faces, which is trained by the Haar-training, with the help of the Intel Open Source Computer Vision Library (OpenCV). Then, we build eye and mouth boundary maps for verifying the face candidates. Lighting compensation technique is used to improve the image quality. Experimental results show successful detections over a wide variety of facial variation in position, color, scale and pose and in in-door/out-door photo collections.