Gross - 4 
descriptions afford important postprocessing computations, such as surface and volume 
evaluations. Furthermore, a large variety of ray-tracing algorithms had been set up for pho 
torealistic visualization. 
The power of B-spline models for shape descriptions in vision and electronic 
photogrammetry is illustrated in figs. 2 and 3 where different image and graphics proces 
sing steps are presented. The goal was to estimate a 3D-shape of a high speed copper beam, 
which was emitted from a shell and recorded via high speed x-ray cameras. Figure 2a 
shows the x-ray image of the beam. Due to some fluctuations of the local histogram, differ 
ent preprocessing steps were necessary to perform edge detection, filling and segmentation 
as appearing in fig. 2b and c, respectively. These methods, however, belong to the standard 
repertoire of image processing and can be found in the corresponding literature [8]. 
Fig. 2 Reconstruction of the 3D shape of a copper beam recorded by a high-speed x-ray 
camera: a) Initial x-ray image. 
b) Edges detected with an anisotropic Canny operator. 
c) Segmented, filled and morphologically postprocessed beam regions with 
medial axis. 
(data source: courtesy Gruppe fiir Rustungsdienste, SA 26, Switzerland, from [9])