摘要： 结合现有的人眼定位方法提出了一种新的算法。该算法首先找出输入图片中所有可能的眼睛点，将这些点按照一定的标准两两结合为可能的眼睛对后，对假定的眼睛对进行模板匹配，选出匹配程度较高的眼睛对。根据眼睛对的距离从图上画出相应的人脸区域，对该区域采用K-L变换，判断其是否为人脸。如果该区域为人脸，就可确定候选眼睛对。

关键词: 眼睛定位, 模板, K-L变换, 人脸识别

Abstract: In this paper, an improved eye locating method is proposed. In the method, possible eye candidates in the image with a complex background are identified by valley field detection. All the eye candidates are grouped and matched with a template. If an eye pair candidate satisfies the criteria, it will be verified by the K-L transformation. Then the verified eye candidates are located as the eye pair precisely.

中图分类号: 

• TP391.41