Abstract: Genetic algorithms are applied to extract the face region in gray-level image with cluttered backgrounds. Based on the geometrical configuration of human face, a simple and efficient face model is designed. According to the grayscale variance of different sub-regions in face, a grayscale distribution model is designed to remove the false face regions. Then the genetic algorithm with strong search ability is proposed to search the input image simultaneously for potential face regions at different scales by calculating the similarity measurement between current regions and two types of face models. Experimental results show that our method is able to detect faces with slight rotation (on-the-plane rotation: -15°~15°, horizontal rotation: -20°~20°, vertical rotation: -15°~15°), and insensitive to facial expression, eyeglasses and mustache. The average computation time for an image of size 320×240 is less than 1 second on PC, and the detection accuracy is about 88%.