关键词: 眼型分类, 眼睑曲线, 眼型指数, 曲线相似度, 最小二乘法, 曲线拟合

Abstract:  Aiming at the problem of human eye classification in facial images, this paper utilizes different types of eyelid and proposes an approach to classify ocular types based on the combination of eyelid contour curve similarity and ocular type index. The description method of contour types is introduced, the sampling points are added on account of obtaining the initial feature points of the eyes; Then on the basis of upper and lower eyelid curves and   inner and outer corner of the eyes, the equation of eyelid contour curve is obtained by the least square fitting method according to the sampling points of the upper and lower eyelids; The slope of upper and lower eyelid sampling points are calculated by the eyelid contour curve equation, the slope similarity of upper and lower eyelid sampling points is described by the normalized cross-correlation coefficient; Finally, ocular type index can be calculated by the eyelid contour curve equation, which combined with slope similarity, can achieve the purpose of ocular types classification. The main contributions of this paper are the definition of upper and lower eyelid contour curves and the similarity between them, and the related ocular type index can be used in ocular type classification. The proposed method reaches an accuracy of 85.17% on the four eye shapes of round, standard, slender, narrow. The experiments prove that the proposed algorithm is easy to implement and can obtain better classification results.

Key words: eye shape classification, eyelid curve, ocular type index, curve similarity, the least square approximation, curve fitting