Information Technology Reference
In-Depth Information
(a) Regular mesh layer
(b) Bound ary layer
(c) Feature point layer
Fig. 7 3D scene rendering using hierarchical decomposition
Table 2 shows the number of feature points, the number of triangles, and the
shape of the reconstructed 3D surface for grid cells according to layers. In the ta-
ble, we consider only four feature point layers and the shape of surface generated
from them is dependent on the location of their depth pixels. When a grid cell has
a regular mesh layer only, we create the 3D surfaces with only 2 triangles. For the
other no-edge-grid cells, of which is represented by the regular mesh and feature
points layers, we generate the 3D surface using from 4 to 10 triangles with from 5
to 8 depth pixels extracted from the regular mesh and feature point layers. For
edge-grid cells represented by quad-tree modes in the boundary layer, we generate
the 3D surface using 20 triangles with 14 depth pixels extracted from the regular
mesh and boundary layers. For edge-grid cells represented by the full modeling
mode, we generate the 3D surface using 44 triangles with 25 depth pixels from a
regular mesh and boundary layers.
Likewise, we can generate a dynamic 3D scene rapidly by assigning regularly-
predefined 3D shape patterns according to layers into the grid cells in each frame. The
generated 3D surface by hierarchical decomposition is covered by the corresponding
Search WWH ::




Custom Search