在人脸图像中确定嘴巴位置的方法

利用迭代式阈值选择算法在人脸的下半区域进行阈值化处理,无需人工干预就可得到嘴巴鼻子等区域,且很好地改进了传统的二值化方法所造成的嘴部轮廓不清晰、连通性不好等缺陷;用腐蚀膨胀算法去掉较小的噪声点,用最小外接矩形近似表示各个连通区域的形状,根据嘴巴形状大小的先验知识确定嘴巴所在的连通区域;用嘴巴连通区域的质心表示嘴巴中心,利用Harris角点检测算法在原灰度图像的嘴巴区域中找到两个嘴角。实验结果表明,本文算法具有较快的速度和较高的精度。

Determine Mouth Position in Face Image

The iterative threshold choosing algorithm is used to changes the gray image into binary image in the low half of a known face image, the area of mouth and nose can be obtained without manual work. This algorithm improves the traditional method limitation, such as vague mouth contour and poor connection quality. The erosion and dilation arithmetic is used to delete the minor noise area; the minimum enclosing rectangle is used to describe the shape of every area; and the mouth area is determined according to the prior knowledge of the mouth shape and size. The mouth center is represented by the centroid of the mouth connection area. Harris corner detection arithmetic is used to detect the two corners of the mouth in the mouth area of the original gray image. Experiment shows that the algorithm presented in this paper is quicker and more accurate than traditional methods.