Face detection algorithm based on double ellipse skin model

In this paper, we develop a novel face detection algorithm for color image in conditions of varying lighting, complex backgrounds and multiple faces. This algorithm relys on a four step process: First, preprocessing the color input image by color bias correction and light compensation and equalization. Second, segmenting and extracting the skin region based on CbCr-CgCr double ellipse skin model. Third, postprocessing the segmented skin image by morphological denoising and facial features holes filling. Finally, locating face region by using AdaBoost cascade classifier and detecting facial features by making use of face geometric and gradient characteristics. Experiment results show that the effectiveness of our algorithm.