嘴部特征提取是人机交互、人脸识别和重建等领域的重要步骤。提出了一种改进的利用变形模板提取嘴部特征的算法。首先对预处理之后的图像用改进的积分投影法确定人脸各特征的水平位置,然后对投影曲线进行分析确定出嘴巴的中心位置,之后利用边缘和灰度信息,对所设计的模板进行全局搜索确定变形模板的最佳参数。算法先提取嘴巴轮廓再根据曲线约束关系判断嘴巴是否张开。实验表明,对于简单背景的人脸图像,该算法可较好地提取嘴部特征,速度快。
Extracting mouth features is a key part in the fields such as man machine interaction,face recognition,face rebuilding and so on.This paper proposes an improved algorithm to extract mouth features using deformable templates. The paper first uses improved integral projection to label the feature.s of the face then analyses the graph of the integral projection to label the center of the mouth,then executes the deformable templates matching in the face picture using only edge and gray energies,to find the optimum of parameters of the deformable templates.The algorithm first extracts the outline of the mouth,then decides the mouth is opening or closed.The test results show that the exact position of the mouth features can be extracted for the simple background facial images by the algorithm quickly.