DMC
fodular
ages, a
-arrays
or look
lencing
) tables
mosaic
posite.
ic. The
; paper
lect the
he last
flow is
e high
ecause
'e field
and up
lightly
rm the
pectral
> same
ds to a
Stefan Diener
along track (X)
cross track (Y)
Figure 1. Sensor footprint on ground.
black quadrangles: panchromatic, grey rectangle: multispectral
2 RADIOMETRIC NORMALISATION OF THE DMC
Even though Charge-Coupled Devices (CCD) have a sensitivity which is quite equably for the whole area of the CCD
plane there is still the need of a radiometric normalisation. The reason for this stems from the accuracy requirement for
photogrammetric applications. Since local changes in radiometry can take effect on the result of an automatic
measurement one has to correct the differences of the sensitivity of single pixels inside a CCD.
But the differences of the resulting gray values do not only stem from the different sensitivities of single pixels. They
also result from other influences like lenses and aperture. And there are more forces left. The DMC has the ability to
compensate the forward motion during an exposure and therefore we have to correct the influences of this TDI (time
delayed integration). Another well known influence on the resulting values of CCD images is the dark current. The
longer the image data stays on the CCD the more this value increases. Thus one has to use the shortest read out time as
possible and one has to correct the influence of this time on the data.
2.1 Classical Normalisation Approach
Many studies on CCD behaviour have shown that the voltage output of a CCD increases linear with the integration time
[Theuwissen95]. Therefore one can use a linear approach to correct the influences shown above.
The classical way to compute the coefficients of the linear model is to take the so called dark and bright image. The
dark image is taken to get the dark current of each pixel of the CCD. Without exposure of the CCD the whole CCD
values are read out and stored in an image. One expects a totally dark image with zeros. Since a CCD has the dark
current we get a dark image but the values inside are not really zero but close to. And the values between the pixels vary
too. The bright image is the second image which is used for the normalisation. For this image one needs a homogenous
source light. That means the light at each pixel of the CCD has the same brightness. This is done by using an Ulbricht
sphere. With these two images one can now compute the linear coefficients (gain and offset) of each pixel using the
following equation:
res
BAM. S——
bright. —dark
res * dark
Offset, m — 9 — —
fet bright, — dark,
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B1. Amsterdam 2000. 83