引用本文: | 吴伟国,闫云雪.人头面部轮廓数学模型的研究[J].哈尔滨工业大学学报,2014,46(11):30.DOI:10.11918/j.issn.0367-6234.2014.11.005 |
| WU Weiguo,YAN Yunxue.Research on mathematical model for human head-face contour[J].Journal of Harbin Institute of Technology,2014,46(11):30.DOI:10.11918/j.issn.0367-6234.2014.11.005 |
|
摘要: |
针对现有图像处理技术中提取人脸轮廓线光滑性差的问题,提出一种人头面部轮廓的分段建模方法. 通过轮廓特征点定义和轮廓分段,采用超椭圆、圆弧和抛物线等曲线建立了人头面部轮廓数学模型,选取100组5种脸型的真实人头面图像经图像处理及优化后,用获得的轮廓样本参数对模型进行验证. 结果表明,对于任一样本,该模型在进行参数优化后都能保证模型平均误差在1.2%以下,最大误差在2.6%以下,验证了该模型的有效性和普适性. 提出了基于该模型的脸型判定方法,给出了各类脸型的模型参数范围. |
关键词: 人头面部轮廓 数学模型 超椭圆曲线 脸型判定方法 |
DOI:10.11918/j.issn.0367-6234.2014.11.005 |
分类号:TP391 |
基金项目:国家高技术研究发展计划(2006AA04Z1,2AA041601). |
|
Research on mathematical model for human head-face contour |
WU Weiguo, YAN Yunxue
|
(School of Mechatronics Engineering,Harbin Institute of Technology,150001 Harbin,China)
|
Abstract: |
Face contour extracted by most existing extraction technology has the problem of non-smooth. Aimed at this situation,an segmentation modeling method for head-face contour is proposed. The piecewise function model is established using hyperelliptic curve,elliptic curve,circular curve and parabolic curve according to points definition and segmentation of contour. Contour sample's parameters obtained from the processing and optimization of 100 real human head-face images which contains five kinds of face verify the effectiveness and universality of the model. The result shows that the average error is under 1.2% and the maximum error is under 2.6% for each sample after optimizing the model. Finally the range of model parameters corresponding to each type of face is given by face shape classification based on the mathematical model. |
Key words: head-face contour mathematical model hyperelliptic curve face shape classification |