Automatic detection localization method of eye feature

In order to extract eye feature accurately, a novel automatic detection localization method of iris and canthus is proposed. First, the face is recognized in a complicated background by using AdaBoost algorithm. Variance integral projection is used to extract eye area. Then, the deformation circular template and the improved optimization matching function are designed to locate the center of the iris and to calculate the iris radius. On this basis, according to the characteristics of the eye structure, both inner and outer canthus direction is calculated using linear template, which is relative to the center of the iris, and then Harris corner detection algorithm is used to set canthus best position along this direction. Finally, this method is validated by IMM face library, and the experimental results show that this method can locate the human iris and inner and outer canthus accurately.