A 3D face model automatic segmentation method for cosmetic surgery was proposed.Combined with multi-resolution models,principal component analysis,nose tip characteristics and intersection curve shapes,a standard frontal and upright position was formed.Then,a group of 3D facial landmarks were detected effectively by use of 2D face alignment technology on 2D image generated by automatical rendering.Finally,a standard anatomical segmentation template designed for different types of cosmetic surgery was automatically fitted on the input model based on corresponding 3D landmarks,and 17 regions were automatically generated.3D face segmentation method was compared to the common curve-fitting hierarchical clustering algorithm,the shape diameter number (SDF) algorithm and SNAKE algorithm.The results show that 3D face model segmentation has better performance than the contrasted algorithms,and under the same experimental conditions,3D face model segmentation is 10,70 and 50 s faster respectively than the three contrasted kinds of algorithms.%提出一种用于美容整形的三维人脸模型自动分区方法.通过不同分辨率模型、主成分分析数据、鼻尖特征和相交曲线形状等进行组合分析,生成标准的正面朝向的三维人脸模型,然后,自动渲染生成二维图像并用可靠的二维人脸特征检测方法生成二维标志点,逆向生成三维人脸标志点.最后,基于对应的三维标志点,将标准制作的解剖学分区模板与输入模型进行自动拟合,自动生成17个区域,用于不同类型的美容整形手术.在实际三维人脸模型分割实验中拟合基元的分层聚类算法、基于“形状直径函数”(SDF)的算法以及SNAKE方法等对比方法.研究结果表明:本文方法所得实际分割效果比对比算法好,在相同实验环境下,本文方法比这3种对比方法分别快10,70和50 s.
展开▼
