IIGH 
mmetry. 
| by new 
1 of the 
vestigate 
n image 
à CMOS 
camera. 
OS-1Ds 
e Canon 
1d 
power 
h more 
)ff-chip. 
es is its 
turation 
€. This 
has less 
/ of the 
| greater 
ceptible 
From a 
se most 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004 
  
of the circuitry is on board. This means that signals have less 
distance to travel and don't have to be piped to other chips on 
the printed circuit board. 
CCDs rely on a process that can leak charge to adjacent pixels 
when the CCD register overflows; thus bright lights “bloom” 
and cause unwanted streaks in the image. CMOS architecture is 
inherently less sensitive to this effect (Table 1). The power 
consumption of CMOS sensors is at least 10 times smaller that 
that of similar CCD sensors and CMOS sensors are more 
durable than CCDs because of the high level of integration on 
the chip. More integration means less external connection that 
are susceptible to corrosion and other problems associated with 
solder joint in harsh environments. It is fairly obvious that 
CCDs offer greater image quality and noise immunity and are 
more flexible from a design perspective. 
CCD | CMOS 
Smallest pixel size | 
Lowest noise 
Single power supply 
Single master clock 
Lowest dark current Low power consumption 
Established technology market base | Smallest system size 
Highest sensitivity | Easy integration of circuity 
Table (1) Comparison between CCD and CMOS technology. 
The EOS-1Ds has made a huge leap in resolution by means of 
the continuous development of the CMOS sensor (Figure 2). 
Canon's CMOS technologies deliver high resolution, low noise 
and low power consumption, allowing photography to make a 
quantum leap in terms of digital image quality. Featuring 11.1 
effective mega pixels the EOS-1Ds provides the highest image 
quality with a digital AF-SLR camera. With its extremely high 
resolution, the EOS-1Ds has been designed to meet the needs of 
those professional photographers who need the ultimate in 
digital image quality. Therefore, it could be ideal for 
professionals active in a wide range of specialties including 
architectural and industrial photogrammetric photography. 
Professional EOS Digital SLR 
Magnesium body, environmentally 
sealed, based on EOS-IV 
Integrated battery compartment / 
vertical hand grip 
11.4 megapixel CMOS sensor 
(primary colour filter) 
Full frame sensor, no field of view 
crop / focal length multiplier 
  
Output image size: 4064 x 2704 or 
2032 x 1352 
Figure 2. General specification of EOS-1Ds (Cannon 2004) 
Various investigations have already been carried out regarding 
the geometric and radiometric evaluation of digital cameras, no 
really detailed examination is known for this CMOS camera. 
Therefore this study investigates the radiometric and geometric 
potential of the EOS-1Ds CMOS camera. 
2. RADIOMETRIC INVESTIGATION 
Image restoration describes the removal or reduction of 
degradations that occurs when the digital image is being 
generated. These degradations include the blurring that can be 
introduced by imaging systems, image motion and the noise of 
electronic or electromagnetic sources. In other words, 
restoration attempts to reconstruct or recover an image that has 
been degraded by using a prior knowledge of the degradations 
(Castleman, 1996; Gonzalez and Woods, 2002). 
2.1 Model of Degradations 
Figure 3 illustrates the process of how a degraded image g(x, y) 
is formed by blurring an ideal image f(x y) with a system 
operation A(x, y) followed by adding noise n(x, y): 
ftx, y) Lh y) | * g(x, y) 
  
n(x, y) 
Figure 3. Model of degradation 
If h(xy) is a linear, position-invariant process, then the 
degraded image is given by: 
g(%, y) = h(x, v)* f(x, y) + n(x, y) (1) 
Here, (*) indicates convolution. 
2.2 Model of Wiener Deconvolution Filter 
For restoration of CMOS images, both the degradation function 
and statistical characteristics of noise should be considered. 
Wiener Deconvolution is based on considering images and 
noise as random processes. The objective is to get an estimate of 
the uncorrupted image is such away that the mean square error 
between degraded and undegraded images is minimized. 
Further, it is assumed that noise and image are uncorrelated, and 
the gray levels in the estimated image are a linear function of 
the levels in the degraded image (Castleman, 1996; Gonzalez 
and Woods, 2002). By these assumptions, the Wiener 
deconvolution filtering in frequency domain reads as follows: 
K(u,v)= 3 gen G(u, v) (3) 
|H (u, v +P, (u,v)! Pr (u,v) 
  
where F(u,v) is the restored image, H(u, v) is degradation 
function (OTF), H *(u,v) is complex conjugate of H(u, v), 
2: ; 
P,(u,v) = IN Qu, v) is power spectrum of the noise and 
2 
P,(u,v) = |F (u,v) is power spectrum of the undegraded 
image. 
An approach used frequently when the quantities of P,(u,v) 
and P,(u,v) can not be estimated is to approximate Equation 3 
by the expression: 
H* Gv) 
Fu,v)= 3 
IH iv) +K 
G(u,v) (4) 
where K is a constant (Gonzalez and Woods, 2002). 
In the follooing, we propose an approach for estimating the 
system degradation model and noise of CMOS images and then, 
by designing a Wiener deconvolution filter, we attempt to 
restore a CMOS image which is corrupted by degradations. 
2.3 Determination of Noise and PSF 
The first step for determining noise and point spread function 
(PSF) of the system is to provide ideal (rectangular) targets, to