Some results are shown here to illustrate the edge-detection and edge-preserving performance of our surface reconstruction algorithm.
Figure 1: The wedding cake example, (a) the original wedding cake
range image and (b) the 10% sampling density. Dots indicate the
sampled data, and the brightness on them indicate the depth. The
brighter the point is, the closer to the viewer it is.
Figure 2: The results obtained with the proposed algorithm for the
10 % wedding cake range data; (a) the discontinuities found and (b)
the reconstructed surface.
Fig 1
shows synthetic range measurements where zero-valued means no
sample data
and the intensity denotes the range. A sampling density of 
of
the range image is used.
Fig 2
shows the discontinuities found and the reconstructed
surface using our algorithm with three resolution levels and
50 iterations in each level.
Despite the sparse data available over a surface with a lot of
discontinuities, our algorithm
gives a good reconstruction.
Figure 3: The second example. (a) the original image and (b)
a 10 % sampling density image.
Figure 4: The results obtained with the proposed algorithm for the
10 % checkerboard image; (a) the discontinuities found and (b)
the reconstructed surface.
In the second example, a checkerbox surface of size
is interpolated.
The sampling density is also 
and the sampled image is shown
in Fig 3. Fig 4
shows the discontinuities found and the reconstructed
surface using our algorithm with three resolution levels and
50 iterations in each level.