基于重采样的三维人脸样本扩充

    3D Face Expansion Based on Face Resample

    • 摘要: 为了对现有三维人脸数据库的数据规模和数据覆盖范围进行扩充,提出一种基于样本重采样的三维人脸样本扩充算法.该方法首先根据面部器官的分布规律对人脸样本进行区域划分,并获得关于各个器官的样本集;然后通过从各个器官样本集中选择器官并重组的方式来获取新的三维人脸样本,为了保证新器官之间的无缝融合,提出了基于薄板样条函数(thin-plate-spline,TPS)的几何信息缝合方法和基于微分算子的纹理缝合方法.实验结果表明:本文提出的样本扩充方法能对现有三维人脸数据库进行扩充;扩充后的样本集可以提高算法的效果.

       

      Abstract: To effectively expansion the amount and data coverage about current 3D face databases,a sample expansion algorithm based on sample resample was presented.First,the raw face samples were divided into patches to form the organ sample sets.Then the new face sample could be constructed by picking up organs from different set and assembling.To guarantee these organs can be seamless merged together,the shape stitching method based on TPS and the texture stitching method based on differential operator were presented.Resultsdemonstrate that the proposed method has good performance on sample expansion,and the effect of algorithm can be improved by the expansion sample set.

       

    /

    返回文章
    返回