Table 1 Specifications of IKONOS satellite and MS imaging
system
Specifications value
Speed on Orbit 7.5 kilometers (4.7
miles) per second
Orbit 98.1 degree, sun
synchronous
Altitude 681 kilometers (423
miles)
Over 8.5 Years
10:30 a.m., local solar
time
3 days at 60 elevation
11 days at 72 elevation
141 days at 89
Operational Life
Descending node time
Revisit time at mid-latitude
elevation
Optical telescope assembly
Assembly mass without the 109 kg
focal plane unit
Total instrument mass 171 kg
Primary mirror aperture 0.70 m
Diameter
Focal length om
Focal ratio 8 14 ;
pixels
Detector array 48x48um
pixel size 4 m (3.2 mat nadir)
Spatial resolution
Spectral range (um) 0.45-0.53, (blue)
0.52-0.61, (green)
0.64-0.72, (red)
0.76-0.86, (NIR)
ADPCM, 2.5 bits/pixel
Data compression technique j
11 bits
Data quantization
2. DARK OBJECT
Among the common atmospheric correction methods, Dark
Object Subtraction (DOS) is a simple and yet useful one to
eliminate the effect of the atmosphere from images, especially
when limited ground information is available (Gebreslasie,
2009). To obtain the dark object, IKONOS should seek zero
reflectance surface covers. In order to achieve zero reflectance
pixels in the green and blue bands, tree shades, and in the red
and infrared bands, water body can provide the dark object.
Also, to avoid the effects of MTFC, the MTFC-off images
should be selected (Pagnutti, 2003).
The smallest value in each band of the image with a cover of
forests and water bodies represents the DN value of dark object
(Soudani, 2006), (Mahiny, 2007). It is often preferred this
approach to radiative transfer model for eliminating the effect of
atmosphere. That is because in the radiative transfer model,
measuring the water vapor in the air along with the aerosols as
well as describing atmospheric conditions along with collecting
earth data is difficult. Nonetheless, from a practical standpoint,
the smallest amount of DN may also contain errors. That is
because selection of the dark object value is based on a visual
examination from the histogram values. Another error in dark
object values as used for atmosphere correction —which is
discussed in this article - is the sensor noise. Despite these
errors, dark object subtraction technique has always been a good
correction for the atmospheric effect on remote sensing data
(Soudani, 2006).
The IKONOS image used in this work is an MS one from Quds
town near Tehran. One of the main reasons for having chosen
this region is that the image has both forests and water bodies.
In Figure l, these scenes are shown in two bands of near
infrared and green. Taking a look at the images, one can realize
that, for example, the body of water in the infrared band and
shade of dense trees in the green band seem dark. The smallest
amount of image pixels in each band is calculated with ENVI
software and shown in Table 2. The small value of infrared
compared to other bands represents a low path radiation and the
high transmittance of atmosphere through the band (Richter,
2006).
3. CCD NOISE MEASUREMENT
The imaging system CCD can usually be demonstrated by three
sub-systems. First, the CCD arrays convert photons within each
pixel into electrons and voltage. Second, the electronic part of
the camera performs a non-linear compression of the voltage
values. Finally, the third sub-system is the analog to digital
converter that generates digital image. During the process of
converting radiation to digital image, various electronic noises
enter into the system. In this paper, we investigate three noises,
i.e. dark current noise, non-uniform pixel noise and Read noise,
the first two of which are from CCD array noises and the last
one is from CCD read circuit noises.
Image obtained from the dark object (Epos) includes electronic
sensor noise in addition to atmosphere effect (E44). The three
noises can be modeled as a theoretical relation as follows:
Figure 1. Separating image into green and infrared bands: the right image represents a body of water in the
near infrared band and the left one shows trees in the red band.
Where
Nnon
that a
2001),
cumul
relatio
out. T
2010).
31 1
This
tempe
challe:
theore
showr
Wher
Here,
To ic
meth:
dark
from
