摘要： 提出一种估计图像中人脸边界的方法。在分析人脸图像特征的基础上，将图像分为5个区域，对每一列求方差，并对构成的方差序列求梯度，估计人脸的左右边界。根据人脸图像的梯度特征和人脸的形状特性，估计人脸的上边界和下边界。实验结果证明了该方法的有效性。

关键词: 人脸边界, 参数估计, 灰度值, 方差, 梯度

Abstract: This paper presents a boundary estimation method of human face images. On the basis of analysis of the face image features, the image is divided into five regions, it computes the variance to image each column. The human face left and right border are estimated based on the gradient to the variance sequence. According to the human face image characteristic and face shape features, the human face upper and lower boundaries are obtained. Experimental results prove the effectiveness of the method.

Key words: face boundary, parameter estimation, gray-scale value, variance, gradient

中图分类号: 

• TP391