Hybrid Scheme for Eye Center Localization in Color Images

Eye center localization is an essential step for the initialization of many face processing techniques like face recognition, facial expression recognition, face pose estimation methods and face tracking. This paper presents a novel technique to localize the eye centers by fusing the geometric and color information that related to the eyes region. The performance of the proposed method is evaluated on XM2VTS database. The validation shows that the eye detector has an overall 98.69% eye detection rate. In addition, experimental results demonstrate great accuracy in challenging cases (wearing glasses) and resilience to pose and illumination variations, achieving significant improvement over other existing methods.