A New Face Localization Method Based on Symmetry Transform and Gaussian-Derivatives
Song Hai · Dianzi xuebao · 2003
In real face recognition systems,face localization is always heavily affected by complex background,uncontrolled lighting condition and low imaging quality.To solve these problems,a novel face localiztion method in natural indoor environment is presented.The main idea is to use the strong symmetry and the three dimensional property of human face.Firstly,edge map of original image is obtained by using the Gaussian derivative operator.Secondly,we introduce some novel rules to the generalized symmetry tranform and radial symmetry transform to localize the glabellum,and then to estimate the scale factor.Finally,accurate eyes localization is realized with similar idea.The results of some experiments are also given which show the robustness and effectiveness of this method.