There is an exploding demand for high fidelity 3D models in a variety of fields, both in the commercial and military sectors. Commercially available 3D rendering engines are able to meet these high fidelity 3D modeling requirements efficiently if they are given bump maps (which modulate the surface normals), as a supplement for the polygonal model. At the present time, however, the development of these bump map textures is a difficult and labor intensive activity. In an effort to reduce the cost of creating bump maps, the feasibility of applying automated techniques from photogrammetry and computer vision to the generation of bump map textures from photographic images will be investigated. A preliminary design will be developed for a software tool that will combine automated techniques with selected user input via a simple user interface. A working prototype will be developed to verify that acceptable levels of fidelity and automation can be achieved. Input from a variety of source data types will be investigated, including photo imagery and existing 3D scene databases. The generation of multiple levels of detail and smooth transitions between bump-maps and high fidelity 3D polygonal models will be addressed.
