A Hybrid Face Detection Approach in Color Images with Complex Background

Human face detection in colour images has been researched extensively over the past decades. Face detection has several applications in areas, such as security access control, visual surveillance, video conferencing, intelligent human-computer interfaces and content-based information retrieval. An ideal face detection system should detect faces from a given image/video regardless of their poses, illumination, scale, age, race, image quality, and image complexity with optimum speed and very low false-positive rate. In this paper a hybrid approach, based on the skin color information and Adaboost-based face detection, is proposed. The key points of the proposed framework are background elimination and down-sizing using the adaptive skin colour classification and segmentation, as well as sub-window size estimation. Therefore, the skin colour segments, as face candidates, were searched instead of the whole image. Meanwhile, the Viola-Jones Adaboost-based face detector was adopted in this research as the final face detector. In order to test the accuracy of the proposed algorithm, the proposed system was implemented and some experiments were also conducted on the standard image datasets such as Caltech (California Institute of Technology) standard image dataset. The proposed hybrid face detection system was compared with the Viola-Jones face detection system. The experiments showed that the proposed approach could efficiently improve the face detection system in both aspects, namely, accuracy (98.88%) and detection time (259.59 ms).