Game Development Reference
observed at the same time, should not have a very large difference of horizontal paral-
laxes and hence of depth. This constraint can sometimes be more demanding than the
basic condition regarding horizontal parallaxes.
Presumably, the distance of the optical axes of the cameras should be equal to the
IPD of the observer. If it is possible and desirable for scenes with “normal'' dimensions,
it cannot be considered in extreme cases:
for stereoscopic aerial views, the distance is several hundreds of metres, to give a
sense of depth, which is not natural but useful;
for a vision of minute objects (insects, for example), the distance is less than the IPD
The selection of this depth is important for stereoscopic vision. We will remember the
following two essential rules:
the greater the distance, the more significant is the sense of depth;
the greater the parallax (which depends on the distance), the more difficult it is
for the observer to achieve fusion of images.
13.1.2 Choice of stereoscopic parameters
The theoretical conditions of creation and of restitution of stereoscopic images make it
possible to choose stereoscopic parameters. For the creation of images, two categories
of solutions were studied: the optical axes of cameras being parallel or convergent. In
both cases, you should follow the psychophysical constraints summarised below:
A The vergence-accommodation relation should be modified a little to avoid
visual strain: the main object observed should have zero parallax (or almost zero).
B The fusion of two images should be possible and easy, involving less ocular
B1 The horizontal parallax should be limited, but large enough to perceive depth.
The limitations of horizontal parallaxes are generally the same in positive and
negative value (1 . 2 ◦ to 1 . 5 ◦ ). But in case of a small screen, objects emerging in
front of it are not recommended.
B2 The vertical parallax should be zero or very limited (20' of arc)
B3 Two neighbouring planes at the display level, observed at the same time,
should not have a big difference in depth.
C To better perceive the dimensions and correct volumes of objects in 3D, it is
necessary for binocular vision (neurophysiological phenomenon) to be in coherence
with monocular vision, especially with the perspectives of objects (cognitive phe-
nomenon). For this, the restored vision should be isomorphous to the vision of the real
world, if this is technically possible. We are talking about orthostereoscopic vision .It
must follow the perception of depth by the laws of perspective and by the laws of binoc-
ular vision. An example is dealt with in detail in section 13.1.5 “Creation of images in
orthostereoscopic vision for a project review''. This involves three conditions:
C1 The distance between the two cameras is equal to the interocular distance.
C2 The stereoscopic images should not be stretched or flattened in depth and
hence the angles of fields of vision on acquisition and on restitution should be