Dense Estimation of Surface Reflectance Property by Combining Multiple Illumination Conditions

摘要

One effective technique for using computer graphics (CG) to represent an object in a real environment is to affix a photographic image onto a geometric shape as a texture. Since this technique is affected by the illumination at the time the photographic image was taken, if the illumination conditions within the virtual and real environments differ significantly, the composition results will be unnatural when computer graphics are used to represent the object. To deal with this problem, a great deal of research has been done on estimating the reflectance properties of objects from range images and surface texture images of objects. However, since an observation of the specular reflection component, which is one of the reflection components, depends on the observation direction, irradiation direction of the light, and shape of the object, various assumptions had to be set for the shape of the object and object surface properties. In this paper, the authors propose a technique for estimating true reflectance properties for objects consisting of various shapes or material properties. Specifically, to observe the reflection components at numerous points on the object surface, they determine illumination conditions related to the irradiation direction of the light based on the shape of the object. In addition, by applying multiple determined reflectance conditions and measurement data to a color reflection model, they estimate the reflectance properties of the object surface at each point on the surface. Through experiments, the authors show that by using computer graphics to compose an object based on the estimated reflectance properties, the object can be faithfully represented for various illumination conditions within a virtual environment.