better than 1%) indicate the problems. Furthermore only 
local slopes of the DSNU and PRNU across the area of 
targets will result in displacements of the target location. 
The total variation of the light fall-off due to the optical 
system and other influences is thus not relevant. Figure 4 
shows the PRNU of a SONY-XC77CE camera with a 
9 mm lens. The light fall-off to the corners and the irreg- 
ularity of the grayvalues are apparent. 
The subtraction of the thermal generated background 
(dark current) via the black reference is not always de- 
signed for highest accuracies. Several different variations 
exist which could potentially lead to very small deficien- 
cies as the amount of dark current is already very small. 
The sensor surface can exhibit global and local deforma- 
tions as well as repetitive surface topography patterns. 
The global deviations have been found to reach 50 um 
(Blouke at al., 1987). The surface topography of the sen- 
sor gates and other functional elements can reach several 
micrometers. The precise effect thereof on the location of 
image features as a function of the incidence angle is dif- 
ficult to asses as the refraction coefficients of the various 
elements vary greatly and the precise location of the sen- 
sitive area of the sensor element depends on a number of 
factors. The displacements are not only a function of the 
radius from the image center but also of the direction, as 
structures exhibit row and column wise regularities. The 
advantages of this surface topography, as compared to 
film un-flatness, is that the displacement is stable in time 
and should thus be easier to determine. Figure 5 shows a 
cross-section of an Interline Transfer sensor. The elevat- 
ed structure of the vertical CCD runs in columns direc- 
tion and can be up to a few micrometer higher than the 
photodiode (light sensitive area). 
Aluminium 
photoshield 
Gate of verti- 
cal CCD 
Photodiode 
  
  
Photodiode Vertical CCD 
Figure 5 Cross-section through Interline Transfer sen- 
sor showing surface topography due to gate 
structure. 
The regularity of the sensor elements can be assessed via 
the performance of the fabrication technology, actual 
measurements of the spacing, and the PRNU. All three 
sources allow to state that the regularity of the sensor ele- 
ment apertures and the spacing is better than 1%. 
A review of sensor technology and their performance can 
be found in Beyer, 1992b and the extensive references 
therein. 
     
2.5 Electronic Components of Camera 
The electronic components of the camera serve to control 
the sensor, to perform analog signal processing, and to 
generate a video signal as well as other (synchronization) 
signals to be transmitted. Cameras with incorporated an- 
alog-to-digital converters, differ only with respect to the 
signal to be transmitted. The assessment of electronic 
components refers to off-the-shelf CCD-cameras pro- 
duced as consumer products or for CCT V-applications 
(Closed Circuit Television). 
  
Preprocessing 
AGC 
CCD sensor LPF 
Clamping 
Gamma correction 
White clipping 
1 Aperture correction 
| 
Blanking 
— | Synchronization 
  
  
  
  
  
  
Timing genera- 
tion 
Sensor control 
  
  
  
  
  
  
Figure 6 Block diagram of a typical CCD camera. 
The analog preprocessing includes numerous signal pro- 
cessing schemes to enhance the signal-to-noise ratio and 
to eliminate electronic disturbances such as clocking 
transients from the sensor charge readout (e.g. correlated 
double sampling). Automatic gain control (AGC) assures 
a constant average image brightness with a feedback 
loop and signal amplification. This feature is helpful for 
human observation and certain applications but can in- 
duce unwanted intensity variations of features due to 
changes in other areas of the image (e.g. change of the 
background of an object). This can also be performed 
with a variation of the exposure time. The low pass filter 
(LPF) eliminates clocking transients and high frequency 
artifacts. They frequently exhibit an asymmetric im- 
pulse-response, reduce the MTF, and lead to ringing and 
a feature shift (e.g. Ddhler, 1987; Lenz, 1988; see investi- 
gation of LPF of frame grabber). Clamping and white 
clipping limit the signal level to the video specifications. 
The analysis of temporal noise as a function of the gray- 
value can indicate another deficiency of a camera. It was 
found for several cameras that saturation is reached be- 
fore the maximum video signal level is attained. Effects 
of gray level shift and a non-linear transfer function are 
only of relevance for radiometric analysis of the imagery 
and should not affect the three-dimensional measurement 
accuracy. Gamma correction compensates for the non- 
linear transfer function of TV-monitors. It destroys the 
excellent linearity of the sensor and should thus not be 
used if a linear transfer function is required. Aperture 
correction serves to improve the MTF, but eventual alias- 
ing components will be enhanced as well. Contour cor- 
rection, similar to an edge filter is included to enhance 
the visual impression, but can lead to degradations which 
are similar to those of the LPF.