Studies On Image-based Approaches For Reconstruction,Compression And Rendering Of Surface Illumination Attributes

Synthesizing highly realistic illuminating appearance is a main goal in computer graphics, and the core of the problem is to find a precise reflectance model for the scene. It is becoming a new research focus to obtain these informations from photographs. The approaches can be commonly grouped into three different categories: inverse rendering, image-based relighting(IBR) and image database methods with static illumination, dense sampling of the light and viewing directions to generate a tabular representation of the Bidirectional Reflectance Distribution Function(BRDF) or Bidirectional Texture Function(BTF). Inverse rendering infers parameters of a reflectance model from observed images. It generally requires a strict environment illumination condition, and can hardly achieve global relighting with real-time performance. IBR needs many sample images under varying lighting directions and it is difficult to reconstruct the high frequncey of the illumination. Recently BTF, which is a representation of real-world textures, has received a lot of attention in computer graphics community, for its ability in modeling surface mesostructures and reflectance variations. The BTF is highdimensional and involves large amounts of data, so progress in this technique to date has mainly focused on data compression, and meanwhile these compression must be suitable for rendering. As a database sampled under local illumination, BTF is short of global effect, which makes it interesting to integrate the BTF with global illumination algorithms such as radiosity. The problems mentioned above are studied in this dissertation, some creative algorithms have been proposed and the highlighted ideas and contributions are described as follows:1) A real-time global relighting algorithm under complicated environment illumination is proposed. Combination of inverse rendering and image-based relighting, the reflectance of objects and a low-order illumination function per face are calculated. To relight the scene under new illumination conditions or viewpoints, the recovered reflectance is applied to the illumination model to get direct lighting, while indirect and ambient ones are obtained from the low-order function. This algorithm...