ul 2004
left
GA
ting
the
e 7;
ble
as-
iter
ere
not
r-
al
a
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B1. Istanbul 2004
0.015 [-.:i- 5:9: 00 0o] amp rou em Aa EP uu e erue ees B parameter A,
E EE parameters A, B.. B,
[| parameters A, A», A, B, B,, QI -cconneen on
[__] Chebycheff polynomials
0.01
0.005 -
0.004 -
0.003 -
0.002 -
0.001
0.0005} >
Maximal Distortion Change
à
E
E
U
3V 3M 4V1 4V2 4V3
Figure 5: Distortion changes in pixels of a normalized image with width=1. From left to right: Logitech camera (1),
Terratec camera (2), HP camera with VGA and 3MB resolution (3), Sony camera with vga and 5MB resolution with 4
different zoom values (4), Kodak DCS 460 without demounting of lens (5S), with demounting of lens (SU). Note that the
abscissa has logarithmic scale.
20
— ate
© e
o
maximal distortion [pixel]
Co
10 7720 Tin A, Hate 10230
aperture angle [degrees]
Figure 7: Distortion (maximal value) of Sony DSC V1 at
different optical (0) and digital (*) zoom levels, resolution:
VGA images.
of changing focus again with the Sony camera, as manu-
ally setting focus to a certain distance is possible with this
camera. However, for this camera the focus setting is not
stored in image headers.
The influence of focus is only visible for small distances,
as to be expected. We found visible changes for distances
shorter than 0.5 m, see figure 8.
For computing radial symmetric distortion parameters out
of two sets of calibration parameters at two focus distances
one can use the relation given by Brown (Brown, 1971).
This relation has been empirically verified with the Sony
camera, see figure 9. Here an optimal fit of Brown's func-
tion for the modeling of the distortion parameters is plot-
ted. For this purpose two "reference" distortion parameters
As] and A,2 were estimated at distances s; = 0.1m and
$9 — 10m.
3.6 Influence of Aperture Settings
The influence of the aperture was only investigated with
Sony camera, because this is the only camera, besides the
N
A
—
e
Principal distance c [pixel]
ho
>
©
Cn
i-re
EE
0.1m 0.2m 0.5m 1m 3m infinite
Focus setting
Figure 8: Principal distance at different focus settings with
error bars 1o and 30 of Sony DSC V1 (zoom 2x).
Kodak camera, with the possibility for changing aperture
values. No influence of the aperture on the calibration pa-
rameters could be observed. The investigation was done
with optical zoom 1x.
3.7 Comparison to results obtained with the Matlab
Calibration Toolbox
We compared results obtained with the camera calibration
toolbox for Matlab of J.-Y. Bouguet, cf. (Bouguet, n.d.)
with ours. This toolbox provides the same camera model
as ours with respect to the basic parameters (principle point
and distance) and the additional parameters A1, A2, As,
B4 and B5, and shear and scale difference, with the option
to select a subset of these parameters. Results are similar.
The achieved accuracies are about a factor of 1.9 worse
wthan ours hen using Matlab calibration toolbox with the
same number of images, which may be expained by the
specific point detector and the degree of robustness of the
estimation procedure. Further inverstigations are neces-
sary.
