Game Development Reference
In-Depth Information
If highly accurate calibration results are required or in cases where low-cost or
wide-angle lenses are employed,
distortion
has to be accounted for. There exist
two ways of modeling the camera distortion, which will be discussed in the
following section.
Camera Parameters
The coefficients of the distortion model, together with the intrinsic and extrinsic
parameters, emulate the imaging process of a camera with very high accuracy.
In addition, the skew parameter could also be included. However, since the
modeled distortion already accounts for the interaction between the
x
and
y
components, the image skew does not need to be considered explicitly, but
instead it is treated implicitly as part of the distortion.
It can be noticed in the previous equations that
f
and
s
x
, respectively
f
and
s
y
, are
always coupled with each other. This means that only the values of
f
x
=
f
/
s
x
and
f
y
=
f
/
s
y
, instead of the actual values of
f
,
s
x
, and
s
y
, are important and can be
recovered.
All these camera parameters, which can be recovered directly from a camera
calibration procedure, can be grouped into a
linear parameter vector
p
l
and a
distortion coefficient vector
p
d
, as follows:
[
]
T
x
w
c
y
w
c
z
w
c
f
f
x
y
p
=
α
β
γ
,
(7)
l
x
y
0
0
[
]
T
p
=
distortion
coefficien
ts
.
(8)
d
These two vectors are further combined into one
camera parameter vector
p
c
as follows:
[
]
T
T
T
.
(9)
p
=
p
p
c
l
d
Search WWH ::
Custom Search