Description of eye figure with small parameters

The individuality of a human face depends on the fine details of the facial components, and it is necessary to extract and to describe these detailed patterns in order to recognize human faces. We propose a method to describe the eye figure with small parameters by classifying their patterns to typical groups. First, an eye image is divided into parts such as eyelid and inner corner, and a set of 1-dimensional slit projections is obtained from the 2-dimensional intensity array. Then, the principal component analysis is applied to these projections to find the major axes which have typical features. The individuality of each eye is parameterized by the principal component scores. The effectiveness of the description is evaluated by generating sketch images based on the parameters extracted from real eye images.