传统的边缘提取算子提取人脸的外轮廓线很难达到理想的效果.文中提出一种基于人脸的几何特征的外轮廓线提取方法:首先采用结合SOBEL和CANNY算子提取边缘信息,然后利用人脸外轮廓特点去掉杂散的冗余边缘和边缘修补.该算法能够较好地提取人脸的轮廓线,而且定位精度高,满足人脸轮廓特征提取的要求.采用本法方法对50多幅人脸图像进行实验,结果验证了方法的有效性.

It is difficult to get ideal facial contour using conventional edge extraction operator. In this paper, a new contour extraction method was presented. It was based on geometrical feature of human face. At first edges were extracted with the combination of SOBEL and CANNY operators. Redundant edges were removed consequently and useful edges were connected using facial geometrical feature. Satisfactory facial contour and good localization are achieved through this approach to meet the requirement for facial feature detection. Over 50 different face images are tested by this method and it is shown to produce results with high accuracy