Eye Location under Varying Illumination
Song Haodon · Journal of Beijing Institute of Graphic Communication · 2014
Face recognition is an advanced biology recognition technology. As an important feature on human face,eye positions are determined to generate normalized face so as to improve face recognition accuracy effectively. This paper implements an eye location to solve the eye location problem under varying illumination. Firstly,we utilize retinex theory to extract illumination invariant. Secondly,an edge histogram descriptor( EHD) is used to obtain eye candidate detection on illumination on normalized image. At last,we employ the support vector machine( SVM) and eye probability map( EPM) to accurately locate eyes. In order to improve computation speed of the method,a multi-level selection strategy which consists of a binocular selection method based on haar-like feature and the EHD is proposed. This approach is able to reduce a plenty of non-eye areas to increase the speed of the SVM. Experiment demonstrates that the proposed method can achieve high detection accuracy and low consumption time at the same time.