A novel pupil localization method based on GaborEye model and radial symmetry operator

The eyes are the most important facial landmarks on the human face. The accuracy of face normalization, which is critical to the performance of the following face analysis steps, depends on the locations of the two eyes, as well as their relatively constant interocular distance. In this paper, we propose a novel GaborEye model for eye localization. Based on the special gray distribution in the eye-and-brow region, a proper Gabor kernel is adaptively chosen to convolute with the face image to highlight the eye-and-brow region, which can be exploited to segment the two pupil regions efficiently. After getting the region of the pupil, a fast radial symmetry operator is used to locate the center of the pupil. Extensive experiments show that the method can accurately locate the pupils, and it is robust to the variations of face poses, expressions, accessories and illuminations.