摘要： 眼睛作为人脸的重要器官，其特征对于人脸表情识别非常重要，因此需要对眼睛轮廓进行提取。该文根据眼睛的轮廓特征知识，提出一种利用变换投影估计形状参数，在形状区域内结合图像信息提取眼睛轮廓的新方法。在分析变形模板和变换投影的特点基础上，对基于变换投影提取眼睛轮廓的方法进行介绍。在人脸数据库JAFFE上的实验表明该方法是可行的。该方法同样可以用于其他形状的检测。

关键词: 变换投影, 扩散投影, 眼睛轮廓, 变形模板

Abstract: As an important apparatus in the face, the eyes features play an important role for facial expression recognition. Therefore it needs to extract the eye-contours. Considering the known contour character, a new eye-contour extraction algorithm is proposed in this paper based on the combination of the shape parameters got from the transform projection and the image information in the eye field. The characters of deformable template and transform projection are analyzed, and then the eye-contour extraction method based on transform projection is presented. Experimental results on JAFFE(a facial image database) show the validity of this method. In addition, the proposed idea can also be used for the detection of other shapes.

Key words: transform projection, diffuse projection, eye-contour, deformable template

中图分类号: 

• TP18