一种改进的三维人脸重建方法

摘要

This paper proposed an improved RBF(radial basis function) deforming algorithm based on the RBF interpolation of scattered data to improve the deficiency of RBF algorithm in 3D face morphable model. Regarding the different characteristics between feature points and non-feature points, different parameter values for the basis function were selected to make the adjustment transformation more reasonable in the model. Then performed a secondary smoothing on the local areas containing feature points using the binary cubic polynomial interpolation method to avoid producing a distortion model,finally it generated an intact 3D face model. The experiment proves that this algorithm can generate the 3D face model faster with higher accuracy.%以径向基函数散乱数据插值算法为基础,针对该算法在三维人脸建模中的不足,提出一种改进的算法:在三维人脸建模中,根据特征点和非特征点的不同特性对基函数选择不同的参数值,使通用人脸模型的调整变换更加合理；然后应用二元三次多项式插值方法对特征点局部区域进行二次平滑,避免了畸变模型的产生,最终得到完整的三维人脸模型.实验结果表明,使用该算法构建的三维人脸模型生成速度快,且具有较高的精确度.