Eye Detection Based on Principal Component Template and Nonlinear Correlation

A novel template matching method is proposed for eye detection. the classic template matching methods directly use images as templates. They are susceptible to variations in scale and light conditions. Moreover, the linear correlation coefficient is used to measure the matching degree without considering the higher-order statistics of images. Unlike the classic template matching, the projection coefficients of principal components are used as templates and the non-linear correlation coefficients for capturing the higher-statistic features is proposed to measure the matching degree. Moreover, the reduction of computation costs is achieved by taking advantage of the eye symmetry.