摘要： 提出了一种基于Snake模型的嘴部特征提取方法,简化了初始化Snake曲线的步骤,从而避免了手工初始化Snake曲线时可能产生的误差。该方法使用块像素变差方法解决图像中的弱边界问题,使用图像能量区域来防止边缘泄漏,得到比可变形模板方法更准确自然的结果。实验结果表明,这种基于Snake模型的人脸特征分割方法具有良好的效果。

关键词: 图像处理, 特征提取, Snake模型

Abstract: A semi-automatic method based on Snake model is proposed in this paper to extract mouth features. The method simplifies the initialization of Snake curve, which avoids the error caused by manual initialization, and a block-pixel variation (BPV) approach is used to resolve the problem of faint boundaries, also an image energy zone approach is presented to prevent boundary leak. Thus, the result obtained by this method is more correct and nature than that of deformable template. The expensive experiments show that the method based on Snake model can locate the position of face features with good results.

中图分类号: 

• TP391.41