Practical BRDF Reconstruction from Real Object using Reliable Regions in Geometry Acquired from Multi-view Images
大野 大志 (1651024)
Recently, three-dimensional (3D) shape reconstruction from multi-view images has become more popular and this is known as photogrammetry. However, its appearance is not plausible enough because the object's reflectance distribution is not reconstructed. In this paper, we present a practical method to reconstruct the bidirectional reflectance distribution function (BRDF) from real object. Our method needs multi-view images that are originally prepared for photogrammetry, and a few additional images for obtaining reflectances. At first, our method reconstructs a 3D geometry using photogrammetry, and then samples a few reflectances to reconstruct BRDF by using the existing analytic method. However, the reconstructed surface normal tends to contain errors and these errors considerably affect BRDF reconstruction. Therefore, we introduce assumptions to extract the reliable regions from the reconstructed geometry and clarify how to determine the regions of optimal light and view directions for BRDF reconstruction. These assumptions are the main contribution of our study. The results demonstrate that our method effectively acquires a plausible BRDF.