计算机与现代化

• 图像处理 • 上一篇    下一篇

基于BSSRDF模型回归人脸皮肤实时渲染

  

  1. (河海大学计算机与信息学院,江苏南京211100)
  • 收稿日期:2019-07-01 出版日期:2019-09-23 发布日期:2019-09-23
  • 作者简介:曹莹(1990-),女,江苏徐州人,硕士研究生,研究方向:计算机图形学,E-mail: g_j_t_o@sina.com; 刘惠义(1961-),男,教授,研究方向:计算机图形学,E-mail: hyliu@hhu.edu.cn; 钱苏斌(1984-),女,博士研究生,研究方向:计算机图形学与虚拟现实,E-mail: 164690050@qq.com。

Real Time Rendering of Human Face Skin Based on BSSRDF Model

  1. (College of Computer and Information, Hohai University, Nanjing 211100, China)
  • Received:2019-07-01 Online:2019-09-23 Published:2019-09-23

摘要: 基于三层材质模型,利用双向表面散射反射分布函数,模拟光线在皮肤表面的散射效果。提出一种采用径向基函数神经网络对次表面散射剖面进行拟合的方法,以实现人脸皮肤真实感实时渲染。使用离线渲染的方式进行预计算,获取训练数据,进行训练,代替传统的渲染方式。实验证明,本文方法能有效减少冗余数据,以较快的速度收敛,并且很好地拟合离线渲染的结果。

关键词: BSSRDF, RBF神经网络, 三层材质模型, 真实感渲染, 实时绘制

Abstract: In this paper, the rendering based on the three-layer material model, and the scattering effect of light on the surface of skin is simulated by bidirectional surface scattering reflectance distribution function (BSSRDF). A approach is proposed for real-time rendering of human skin, which uses the Radial Basis Function neural network to fit the subsurface scattering profile. The training data for neural network are precomputed by offline render. Experimental results show that the approach is capable of reducing the redundant training data, evaluating fast and rendering well.

Key words: BSSRDF, RBF neural network, three-layer material model, realistic rendering, real-time rendering

中图分类号: