Realistic Relighting of 2D Facial Images Using Ratio Optimization

Abstract

In this paper, techniques of estimating the direction of light in 2D facial images and realistic reconstruction method are proposed. The irradiance environment map of the input image is approximated using the spherical harmonics(SH) lighting basis function. Using the ratio between the SH coefficient of the reference data set and the SH coefficient of the input image, the optimal direction of light is estimated. Then, by applying lighting functions, the reconstruction is can be made on the 2D facial image.