a strong
parison of
> original
nts have
e manual
a random
After the
sed on an
its, which
nd 2.7um
e of the
nt on the
| the rod,
duced to
ize again
(figure 6)
urse the
ement of
of such a
analytical
ic errors,
direction
between
neter 14.
error to
4. RADIOMETRIC CHARACTERISTICS
In general the fast CCD-linear arrays and CCD-arrays do
have a limited dynamic range in relation to the photos.
The Rastermaster is using a depth of 8 bit corresponding
to 256 gray values. This is usually sufficient. The relation
between the gray values and the optical density of the
film has been determined by means of a calibrated Kodak
gray scale. The whole range of the scanner settings, that
means speed and diaphragm, has been checked. Figure
7 gives an overview over some of the results. Caused by
the different settings the relation between the gray values
and the optical density is primarily shifted in the direction
of the optical density. The dynamic range of
approximately 1.5D is not changed by this.
satisfying results have been achieved. In the meantime
another light source is available with a 5 times enlarged
orightness.
400,00
350,00
300,00 |
ë 250,00 |
= 200001
2 150,00
% 100,00 {
50,00
0,00 ;
e e e Le e © e e
o o e o e o o o
e e ce e e e e e
e eo © eo e e e e
x wn © > oo Oo e =
4 b v ve
wavelength [nm]
blue
Figure 8: spectral sensitivity of the sensor
Of course in a usual aerial positive color film the blue
component is not very intensive, but this
also should not be lost. More important is
300,00
speed 220, diaphragm 7
250,00 | Pr
speed 400, diaphragm 7
200,00 |
9 speed 320, diaphragm 8
S 150,00 | speed 320, diaphragm 9
2
: speed 320, diaphragm 11
100,00 |
50,00 |
0,00
the blue color in the case of false color
images. With the new color line sensors
and an improved illumination the problem is
reduced, but it is still existing.
Due to the fact of a limited density range,
the optimal selection of the scanner
settings is important. An over-saturation of
the CCD-elements has to be avoided
because it can cause a very strong
blooming effect. In the extreme case the
whole sensor line can get blind, that
means a larger area will have the gray
value 255. By this reason we cover the
E EM t v n di m c
Saat Wid, © w ornv was X
e. Ce e. eo e © © c € v
Figure 7: relation gray
value - optical density
1,58 |
optical density
1,74 |
area around the scanned photo with
cardboard. The optimal results are
achieved if the bright parts of the photos
are just getting the gray values 255, that
1,89
2,04
The relation between the gray values and the optical
density is far away from a linear function. Of course it is
possible to linearize it with a look up table but this has
only an advantage for the visual interpretation and the
disadvantage of a loss of information in the bright part. If
the digital image shall be used for an image correlation,
better results can be reached with the original gray
values. The separation of elements with different optical
density in the dark image areas (larger optical density) is
limited, that means it is difficult to scan photo negatives.
In the case of photo positives the important information is
included in the bright parts, here we do have a very good
separation but information in the shadows are reduced. It
is also not a problem to scan positive color film in the
panchromatic mode. Here we do have the same relation
like with photo positives.
If a color image shall be scanned in color, this has to be
done three times with the computer controlled filter wheel.
It is enlarging the scanning time at least by the factor of 3.
Especially with the blue spectral range there are some
problems. The sensitivity of the used sensor is much less
in the blue range (~400 - 490nm) and with the light source
of the RM1 installed in the University of Hannover no
75
means there is a loss of information in the
dark parts of the images. Usually this is not causing
problems in the case of positive black and white or color
film.
The empirical determination of the optimal scanner
settings is very time consuming. It takes approximately 30
minutes with the RM1 and it has to be done at least for
every photo flight. Within a photo flight the settings only
have to be changed if a general change of the landscape
is within the area. As mentioned before, the optimal
settings are shifting the bright parts of the photos just to
the gray value 255 to avoid an over-saturation. It is much
more fast to scan the photo transparencies with a desktop
scanner like the HP Scanjet IICX. In the Institute of
Photogrammetry and Engineering surveys we only do
have the possibility to use the desktop scanner with
reflected light, that means the scanned transparencies
are not looking very nice. The density range of the HP
Scanjet is also smaller in relation to the RM1 (see figure
9) and the gray values cannot be compared directly, but
the density of the bright parts can be measured and with
an empirically developed table the optimal RM1-settings
are available within one minute.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B1. Vienna 1996
