| dc.description.abstract | As computers can only represent and process discrete data, informationgathered from the real world always has to be sampled. While it isnowadays possible to sample many signals accurately and thus generatehigh-quality reconstructions (for example of images and audio data),accurately and densely sampling 3D geometry is still a challenge. Thesignal samples may be corrupted by noise and outliers, and contain largeholes due to occlusions. These issues become even more pronounced whenalso considering the temporal domain. Because of this, developing methodsfor accurate reconstruction of shapes from a sparse set of discrete datais an important aspect of the computer graphics processing pipeline. In this thesis we propose novel approaches to including semantic knowledgeinto reconstruction processes using template based shape processing. Weformulate shape reconstruction as a deformable template fitting process,where we try to fit a given template model to the sampled data. Thisapproach allows us to present novel solutions to several fundamentalproblems in the area of shape reconstruction. We address static problemslike constrained texture mapping and semantically meaningful hole-fillingin surface reconstruction from 3D scans, temporal problems such as meshbased performance capture, and finally dynamic problems like theestimation of physically based material parameters of animated templates. | en_US |
