|Title||Implementierung und Analyse GPU gestützter 3D-Rekonstruktion|
|Tutor||M.Sc. Kai Ide|
|Professor||Dr.-Ing. Thomas Sikora|
|Abstract||The demand for real time 3D reconstruction, that is able to handle dynamic content too, rises steadily. The systems must be able to reliably merge multiple partial views to one high-quality model for a satisfying result when displayed.|
This thesis reviews five individual methods from the area of computer vision/graphics with the object to improve the spatial and temporal consistency of the reconstructed views. The views stem from a GPU supported phase shift 3D reconstruction.
The observations of the single methods lead into the development of two additional procedures which combine the single methods to maximize the spatial consistency. The temporal consistency of the reconstructed views is improved with the help of multiple filters, concluding with 3D models of increased quality.
|Key words||implementation, GPU, 3D reconstruction, computer vision, computer graphics, multiple filters, phase shift|