PI Diagram Based Face Detection with AdaBoost in Color Image

This paper proposes a novel method for detecting faces in color images using AdaBoost algorithm combined with PI diagram. First, potential face area is detected using PI diagram which is constructed by the principal vector of image´s RGB vectors and illumination intensity. Then, the full areas are gotten by the growth algorithm. Finally, these face candidates are scanned by the cascade classifier based on AdaBoost for more accurate face detection. Experiments show that our approach is accurate than AdaBoost and has lower error detected rate than Skin color method.