Game Development Reference
each zone (see figure 11.4). But it is also possible to distribute the rendering of a 3D
scene on other criteria:
Temporal, where each machine computes only one image on n, n being the number
of machines in the cluster
Anti-aliasing, where each machine computes the image following the same point
of view, but slightly offset (by a fraction of pixel).
In theory, there are some other types of distribution, but they are rarely used in
Depth, where the 3D objects are computed by different machines depending on
their position in the z axis (distance from the point of view).
Geometric, where each machine is in charge of computing a certain group of 3D
objects (plants, walls, persons, etc.).
All these methods can of course be combined, provided that sufficient number of
machines are available.
Solutions dedicated to distribution of rendering
It should be noted that some distributions like splitting the image into zones or distri-
bution of stereoscopy can be easily carried out without requiring additional electronic
hardware, by connecting at the most one machine per projector. The final resetting
of images can be carried out, if need be, by a simple superimposition of the projected
images. On the other hand, if we want to increase the number of machines per projector
(more than one) to create a more efficient system, a hardware solution for resetting the
images would be a must. DVG technology is an interesting solution marketed by Orad;
it offers an electronic card connected to a standard graphic card which makes it pos-
sible to combine multiple video flows. This thus helps to reconstitute the images from
each of the elements distributed and computed by different machines of the cluster.
Calibration of an immersive system with projection on large screens is a crucial point.
The most efficient system in the world will be completely unusable if it is not calibrated
correctly. The three most important parameters to be taken into consideration are the
calibration of the geometry of images, uniformity of colours and brightness. All along
the graphic chain of production of an image, from the system capturing the point of
view of the user (tracking) to the perception of the image projected on the screen, there
are numerous causes of potential errors and inaccuracies. Only a detailed calibration of
different parts of this graphic chain can help in controlling or minimising these errors.
Capturing the point of view (tracking)
In systems with multiple screens, the user's point of view is rarely considered to be
fixed, which makes the use of a point of view tracker redundant. In a standard point
of view tracking system, the tracker must ensure that the pair of stereoscopic images,
which is to be displayed, is coherent with the actual point of view of the user: the
points of view of the two images must correspond to the optical centres of the two
eyes of the user. Since it is impossible to have trackers in place of eyes, it is important