Game Development Reference
Left image plane
Right image plane
Figure 13.8 Positioning of two fictitious or real cameras with converging optical axes
Figure 13.9 The vertical parallaxes are exaggerated in the figure
some installations where two observers have their head tracked, the display is theoreti-
cally planned for a single observer at a defined position. In practice, these pseudoscopic
movements are troublesome in special cases. For example, in teleoperation, the oper-
ator may need to move by controlling his manipulator and risks getting disturbed
visually by these false movements.
What should the criteria for positioning the two fictitious or real cameras be based
on? Offhand, you can think of making the two optical axes converge towards the main
zone to be observed. In this case, the projections on the two planes give images with
horizontal and vertical parallaxes (these images correspond to two images displayed
on the screen, to a nearest multiplicative factor (Figure 13.8).
But this case is not so desirable because it involves vertical parallaxes that make
fusion of images difficult. For example, the right and left projections of the front side
of a cube centred on the axis Zr give the following images in Figure 13.9.
Studies have shown that merging is difficult for vertical parallaxes of an angle
greater than 20' of arc (Julesz, 1971). It is therefore desirable that the two optical axes
are parallel to avoid vertical parallaxes that cause problems to the display (Baker, 1987;
MacAllister & Hodges, 1990). The algorithms that make the creation of computer-
generated stereoscopic images possible, in this case, are relatively simple and quick
(calculations of translations and projections).
Another rule regarding horizontal parallaxes of images is to be considered to facil-
itate easy vision. To be merged easily, two neighbouring objects at the display level,