Home  |  Centre for Mathematical Sciences  |  LTH  |  LU
Title: Face modelling - 3D object tracking and segmentation
Authors: Lading, Brian
Year: 2008
Document Type:Licenciat Thesis
Status: In Press
Refereed: No
Publisher: Centre for Mathematical Sciences, Lund University
BibTeX item:BibTeX
Extra: matlth, matlth-vision
Abstract: This thesis investigates the design, the construction and the performance of an image segmentation system, capable of segmenting most classes of 3D shapes in 2D images. With focus on a well known deformable surface, namely the human face, the steps from data collection and handling, through shape registration and analysis, to 2D image rendering and segmentation are explored. Novel approaches are presented for the registration of deformed 3D surfaces and in segmentation of shapes in images. The registration is done by mapping the data into geodesic distance space, where the deformed surfaces are less expressive than in Euclidean space, and optimal segmentations are estimated by automatic construction of artificial active appearance models. The system is implemented in open source programs and developed tools plus the full rendering, modelling and segmentation system is made available to the public.

 

Back

 

Questions: webmaster
Last updated: 2018-08-02

Centre for Mathematical Sciences, Box 118, SE-22100, Lund. Phone: 046-222 00 00