Robust eye extraction using deformable template and feature tracking ability

A new method of eye extraction from facial image using deformable template is presented in this paper. In the new method, we makes full use of prior knowledge of the eye by employing the tracking ability, which is a good feature measurement proposed by Tomasi (April 1991) for selecting features for tracking, in the energy function. With this new proposed energy term, a special energy function is designed to locate the iris, and the eye template would be attracted to their real location more accurately. In order to let the new energy term react on the image, a three-epochs minimizing process is also applied. Some experimental results show that the new method works well.