Adaptive Lighting Modeling in 3D Reconstruction with Illumination Properties Recovery

In 3D reconstruction, appearance is always overlooked compared to geometry and motion. The existing methods prefer texture-based shading, where the reconstructed appearance cannot perform re-lighting, thus weakens the authenticity, aesthetic and availability. Although excellent attempts were made in some works at lighting model-based shading in 3D reconstruction, still they fall short of differentiated rendering of materials. In this paper, we propose a method that can make an artistic effect on the recovery of illumination properties in 3D reconstruction from a single RGB-D image. Our method builds adaptive lighting models for different surfaces, which are used in re-lighting and recovering surface properties. We start with the establishment of coordinate system conversion tables for rapidly matching surface parameters and location. Then, a projection-based triangulation method is developed to generate polygonal mesh. After that, we propose a physically based lighting model with few variable parameters to recover surface illumination properties, in order to accurately describe the reflection of light from different materials. Furthermore, we develop an adaptive algorithm for generating parameters. Since the algorithm is based on the surface consistency of albedo over channels and training-less also, it is applicable to a surface of any shape and material, and the physically based lighting model brings a more vivid re-lighting. The obtained experimental results demonstrate that our method could achieve an excellent reconstructed appearance in both artistic view and verisimilitude.