Graphics Reference
In-Depth Information
Figure 10.11.
Handling occlusions
with form factors.
Figure 10.12.
The geometry for rectan-
gular form factors.
z
D
F
Q =
D
A
.
22 2
(
)
p
yzl
++
Finally, as mentioned earlier, the biggest cost in the radiosity method is in com-
puting the form factors. Precomputing them helps, but there is a storage problem.
The form factor matrix is basically an n 2 matrix. Since many patches cannot see each
other, many entries of this matrix are zero, but many nonzero entries remain. An esti-
mate from [CCWG88] indicates that one would need on the order of a gigabyte of
memory to store the form factors with 50,000 patches. Using the progressive refine-
ment algorithm this n 2 storage requirement is eliminated.
Aliasing also rears its ugly head with radiosity methods. The biggest culprit is the
uniform subdivision used in the hemicube method. One approach ([WaCG87]) is to
change the orientation of the hemicube with respect to the surface normal in a
random way as one moves from patch to patch.
Being view independent, radiosity methods separate the shading problem from
the visible surface determination problem. Because the basic radiosity method does
not handle the specular part of light very well, something that ray tracing does, there
Search WWH ::




Custom Search