Obtaining Shading Properties from Multi-Viewpoint Images

摘要

Integrating virtual models into real environments requires realistic lighting and shading for believable results. Current methods of obtaining this realism require extensive effort on the part of the user or highly specialized equipment. We propose a method to obtain the shading properties of an object by using only a short film or set of pictures from the user centered on a specific object. We reconstruct the object from the images and fit the spherical harmonic basis functions to the luminance values with respect to the surface normals of the object to obtain the lighting properties of the environment. The results show that our method is able to create plausible synthetic images with realistically shaded virtual models. Our method can be applied to consumer augmented reality services because it only requires images that can be captured by devices such consumer-level cameras.