Image-Based Capture and Rendering with Applications to Urban Planning

Overview & Synopsis

This thesis addresses the capture real objects to be integrated in a virtual environment and their efficient display. The thesis has two parts; in the first part, we present two new methods for the capture and rendering real objects using image-based techniques, and in the second part, we present an complete virtual reality system that allows the use of different realistic techniques for the design of urban projects, as well as a first evaluation of this system. The first method proposed allows the capture of a real scene with relatively simple geometry, for example building façades. The capture is done using an small number of photographs. With this method the captured objects have a compact representation; the work of the computer graphics artist is simplified to allow higher quality render and the rendering method does not need geometric subdivision as in previous methods. The second method proposed is for the capture of complex objects, in our case trees, using as input a series of photographs taken around the object. A first estimation of opacities is done with a method inspired by tomography, that give us a volumetric representation of the tree. The high frequency detail of the leaves is added by an algorithm of texture generation using the original images. The result is a very realistic rendering of real captured trees, that can be added in a virtual environment. The virtual reality system presented in the second part of the thesis integrates image-based rendering methods, and other algorithms for the realistic rendering (shadows, 3D sound, etc.). The multi platform system works on systems such as the Workbench of the INRIA Sophia-Antipolis or the RealityCenter of the CSTB. Our work on this real world project, the construction of the Tramway of Nice (France), allowed us to evaluate the choices on realism and the interface of the system. The architects in charge of the Tramway project on the Garibaldi square in Nice performed a set of test, and we present here the first conclusions of our observation.