Face Detection Algorithm with Multi-stage Decision Based on Skin-color and Geometrical Features

In this paper the face detection algorithms in natural images are deeply studied, and a novel algorithm with multistage decision based on skin color and geometrical features is proposed to handle the shortcomings of high false detection rate and long detection time in traditional algorithms based on skin color, template or Adaboost. This algorithm can effectively decrease false detection rate and detection time. The method is described as follows. First, the original image is preprocessed by scaling operation and adaptive lighting compensation. Then, the skin color model is established according to skin color feature, and the face candidate regions are generated after morphological operation. Finally, facial geometrical features and Adaboost algorithm are utilized to realize multi-stage decision, and the decision is used to detect and locate the face regions. The experimental results show that the face detection rate can achieve 94.52% and the proposed algorithm has lower false detection rate that compares with traditional algorithms, and faster detection speed at the same time for image in presence of complex background as well as multi-target input.