Faces detection using skin color, region-props, bounding-box and neural networks system

Abstract

Face detection is an important prior step for face recognition system which is widely used in security systems, face verification systems, telecommunication, video surveillance, facial expressions recognition, status authentication, etc. The proposed system is applied on many images which contain persons and extract the faces out of there automatically. The skin color, region-props and bounding-box are used as preprocessing tools for extracting regions. The Neural Networks (NNT) are used to recognize these regions are faces or not. This system has the ability to detect faces with various image conditions: Different poses and facial expirations, different and complex backgrounds and different luminance and lighting conditions. The system is tested on several color images. The detection rates for used databases are 67.6% for images which has background nonluminous, 88% for very dense images and 100% for non-dense images with a luminous background.