Search

Illumination for mixed reality consists of simulating the illumination of 3D scenes composed with real and virtual 3D objects. Current research into illumination methods for mixed reality focuses on im proving the quality of the simulated illumination, while assuming that the required input data, the scene geometry and radiance, can be accurately acquired. This thesis provides methods that reduce the impor tance of this condition. As a first contribution a classification of the illumination methods for mixed reality is presented, from which can be derived that the existing methods are not robust against inaccurate input data. This limits the applicability of the previous methods to easy-to-model scenes. In complex-to-model environments, the capture of the geometry and radiance can be compromised due to camera movement, object movement, and illumination changes. Examples of such uncontrollable environments are scenes with natural lighting and deformable objects. The influence of an incorrect geometric model and radiance distribution on the quality of the illumination methods is analysed. The problems are examined and new solutions are proposed to enable illumination simulation for mixed reality of complex-to-model scenes. Rather than improving the geometry and radiance capture using expensive equipment, this thesis provides low-cost solutions. Two solutions are separately proposed to compensate for the inaccurate geometry and for the radiance capture, while a third solution embeds the consideration of inaccurate geometry and the instability of radiance information into one single method. The combination of the presented methods allows the simulation of consistent illumination for mixed reality without the need of increasing the efforts to capture more accurately the geometry and radiance. Making illumination methods for mixed reality more accessible and applicable to real-world environ ments is the key contribution of this work to the state of the art in illumination simulation for mixed reality.