基于PS-Level Set的嘴唇几何形状定位模型

针对面向唇读的水平集模型在嘴唇分割中存在边界过收敛和过早收敛的问题,文中提出了一种改进的基于先验知识的水平集模型(简称为PS-Level Set)来进行嘴唇几何形状的定位.PS-Level Set模型利用改进的差值能量函数引入嘴唇形状的先验信息.在曲线演化过程中,反复比较演化曲线和先验曲线的差距,使曲线的演化形状逐渐逼近先验模型形状,从而更精确地收敛于目标物体实际轮廓.实验表明,用PS-Level Set模型定位嘴唇几何形状的准确率比用水平集模型提高了8.38%.

Detection Model of Geometric Shape of Lip Based on PS-Level Set

Level set model is likely to pass through weak lip boundary or may not able to pass some narrow parts of the object. To resolve above problems, we proposed that a new level set model based on Prior Shape (PS-level set) for lip geometric shape detection in this paper. PS-level set model incorporated prior shape information into a new shape energy function. During the evolution processing of the curve, this model repeatedly compared the difference between the curve and the prior shape curve, and make the evolution curve shape approached the prior shape. So that the evolution shape could convergent the target objects more accurately. Experiments on mouth images show that the accuracy improvement of the lip contour detection using PS-level set model is 8.38% over level set model.