Game Development Reference
In-Depth Information
with the simple shape factor:
1
A
1
A i
F j =
K ( x , x ) dAdA =
K ( x , x ) dA i dA j
(15.18)
x
x
A i
A j
The method of radiosity has been improved several times after its introduction in image
synthesis. The hierarchical resolution (Hanrahan et al., 1991) decreases the complex-
ity by solving it at different levels, which gives a multiresolution approach using the
wavelets functions for the projection of radiosity (Gortler et al., 1993b). In fact, when
the surfaces are distant, a rough subdivision is sufficient to obtain good results, whereas
when the surfaces are closer, a finer subdivision becomes necessary. Smits et al. (1994)
and Sillion (1994) suggest an extension of this work by clubbing the surfaces together
into groups (cluster). The energy transfers between groups or between groups and
surfaces can be calculated directly as per the required precision. The question of how
to effectively create the groups of surfaces or objects still remains unanswered. The
Monte Carlo method was also used to solve the radiosity equation in a different man-
ner (Bekaert, 1999). To conclude we can say that the integration of more complex
BRDF models is possible (Immel et al., 1986; Sillion et al., 1991; Gortler et al., 1993b;
Christensen et al., 1996; Stamminger et al., 1998), but simple glossy curved surfaces
are not managed correctly. A complete and detailed panorama of radiosity methods is
available in (Cohen & Wallace, 1993; Sillion & Puech, 1994; Stamminger, 2000).
The algorithms of radiosity, though very complex to implement, are very effective
in terms of image quality if the scene is static. The result of the simulation, obtained by
pre-processing in the scene space and stored directly on the meshes or in the form of 2D
textures, makes it possible to develop walkthrough type virtual reality applications in
realistic environments. Architectural applications are a good example of virtual reality
applications where a radiosity approach can be used. If the dynamics of the scene and
the possibilities of interaction are limited to some well identified objects, it is all the
more possible to combine a simulation of the lighting by radiosity for the static part
of the scene with a real time simulation by local illumination on the dynamic objects.
15.2.2.2 Local Illumination and virtual reality
Though the quality of images produced by a global illumination method is very inter-
esting for virtual reality considering the visual immersion it offers, it is not accessible
for general applications at the moment. The calculation cost of such methods of light-
ing and the physical realism necessary for the description of the scene are the two main
obstacles in using global illumination methods. Besides, several approaches that are
further localised have been developed. These methods are mainly based on a percep-
tual approach of rendering in virtual reality and try to calculate only the information
necessary for understanding the scene. In this manner, thesemethods are often “image''-
based, both in terms of calculation space and representation of simulation data.
We can derive a simplified expression suitable for real-time rendering on the basis
of the general equation of rendering 15.7. Several assumptions are made for this
purpose:
Point light sources and in limited numbers;
Sampled spectrum of Red, Green, Blue;
 
Search Nedrilad ::




Custom Search