Matching and fusing 3D-polygonal approximations for model-generation

Automatic generation of 3D-wireframe models from samples of images requires methods for data integration. Input is a sequence of 3D-segmentation data from different viewpoints of one single object and a coarse estimation of the movement of the viewpoint from one view to the next. Segmentation data consists of 3D-polygonal curves used as an approximation for arbitrary (non-closed) 3D-curves: only the ends of the polygonal curves are assumed to be stable, the position of all other points of the polygonal curve may change. The model-generation approach matches data of subsequent views and fuses matched data. The matching algorithm is able to overcome problems of varying polygonal approximations of the same 3D-curve. The matching-and fusing algorithms, including experimental results, are presented