Parametric Model of MPII Building

Dean's award for outstanding master thesis

The scope of this master thesis is to study predictive image synthesis. Predictive image synthesis is a part of photo-realistic rendering, which is based on using physically exact algorithms for computing global illumination. The prediction of the final images is based on a real world model, thus the results can be objectively evaluated. The aim of this work was to extend the previously created 3D model of the MPII building, which consisted of building geometry, approximated materials and measured light characteristics. This static model of the MPII building was parameterized, allowing the following dynamic modifications of the model: i) geometry modifications of movable elements such as doors, windows or window shutters, ii) controlling the lights in the building, iii) modifying the cameras in the virtual scene. Further, a user interface was implemented, which allowed model animations and generation of input data for a physically based rendering algorithm. The correctness of the implemented parametric model of the MPII building was verified by creating five reference video sequences of the animated model.