基于弹性模板的嘴巴轮廓提取

Mouth Feature Extraction Based on Deformable Template

本文提出了一种改进的更符合唇形的嘴巴弹性模板，结合一组利用了原图像灰度信息，边缘及谷值区域信息的能量函数，对微张或抿紧的嘴巴都能使模板与其很好匹配．此外，本文提出的保守的能量下降算法及归一化能量函数简化了经验参数的选择．实验结果证实了本方法的正确性及有效性．

This paper proposes an improved mouth deformable template which can fit the shape of the mouth more correctly. Combined with a set of energy functions defined by image intensity, edges and valleys in the image intensity, the template can match the open or closed mouth and extract its contour. In addition, this approach diminishes the difficulty in selecting empirical coefficients by our improved energy functions and conservative energy descent algorithm (CEDA). Experimental results clarify its efficiency and superiority.