Game Development Reference
In-Depth Information
View transformation
The objection of the view transformation is to reconstruct a virtual view
V
D
for
a virtual camera
C
D
from a pair of stereo views
V
L
and
V
R
, which are generated
from two cameras,
C
L
and
C
R
, respectively.
As a starting point for the following discussion, without loss of generality, the
WCS can be selected such that:
[
]
[
]
[
]
T
T
T
,
,
,
t
=
1
0
0
t
=
−
1
0
0
t
=
x
y
z
cL
cR
cD
cD
cD
cD
where
t
cL
,
t
cR
, and
t
cD
are the position vectors of
C
L
,
C
R
, and
C
D
, respectively. This
means that the x-axis of the WCS lies on the baseline
b
of
C
L
and
C
R
, and points
from
C
R
to
C
L
. The origin of the WCS is at the middle point on
b
, that is, the unit
of distance is
b
/ 2.
In the general case, the view transformation process can be divided into five
steps (see Figure 7):
1)
Rectification
: Transforming the stereo views
V
L
and
V
R
into a pair of new
views
V
rectiL
and
V
rectiR
, respectively. The two virtual cameras
C
rectiL
and
C
rectiR
, which generate these two new views, are parallel to each other and
Figure 7. The view transformation framework. Multiple separate steps
together eliminate three major differences between the final novel view V
D
and the two original views V
L
and V
R
: 1) Photometric differences, such as
focal length, aspect ratio, etc.; 2) Position in 3-D space (x, y, z); 3)
Orientation.
Search WWH ::
Custom Search