The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bl. Beijing 2008
1350
GCPs were collected by a RTK GPS survey assisted by the
RESNAP-GPS permanent network of the Lazio Region,
(managed by the Area di Geodesia e Geomatica - Sapienza
Università di Roma) with a 3D accuracy of 15 cm. The block
triangulation was performed both by Leica Photogrammetric
Suite (LPS) 9.1 and BLUH software. Finally, the image
matching and the DSM generation at 2m grid spacing were
performed using LPS.
Figure 1. Overview of Cartosat-1 Band A. The image size is
12’000 x 12’000 pixels. Some crater lakes are visible.
3 RADIOMETRY ANALYSIS
Within the chain from image sensing to the final value-added
product the quality of the images plays a crucial role. Image
quality is defined by several parameters, as the radiometric
resolution and its accuracy, represented by the noise level, and
the geometrical resolution and sharpness, described by the
Modulation Transfer Function (MTF). In the next sections these
parameters are investigated for the scenes used in this work.
Other radiometric problems, like vertical striping effects, have
not been observed at a significant level.
3.1 Image noise analysis
The image noise characteristics are important for the image
matching process in the DSM generation.
Nowadays, most of the linear array sensors have the ability to
provide more than 8-bit/pixel digital images. The Cartosat-1
sensor provides images with 10 bit/pixel, that means 1024
available grey levels, but 99% of the original pixel values vary
between 0 and 255.
A preliminary analysis was carried out to investigate the noise
dependency on image intensity. According to the method
proposed in (Baltsavias et al., 2001), and (Zhang, 2005) the
noise characteristics of the stereopair were analyzed using the
standard deviations of the grey values in non-homogeneous
image regions, that allow to evaluate the noise variations as a
function of intensity.
In order to evaluate the noise level, a window (3x3 pixels) is
moved within the area by 3-pixel steps in both directions and
the standard deviation and the mean grey value are calculated
for each window. The grey level range is divided in bins, and
the standard deviations are assigned to a bin according to the
mean grey value of each window. In each bin the noise is
estimated as the mean of the 5% smallest standard deviations,
under the hypothesis that the variability within the window
should represent just the noise and not the different texture. The
results reported in Figure 2 indicate that the noise is intensity
dependent showing smaller noise values for the bins of low
grey levels than for the high grey ones. This is a well known
effect, also present in digital aerial images.
NOISE
CNTfCDCOOCM’^-lOT-rv.cO
C0CDO5CMCDO>CNLOt— <OCM
COCOr^-Ov-COlOCOCNi
oiCNicooicqmx-co
T- T- T- CM CM in <o
BINS
Figure 2. Cartosat-1 noise level estimation.
3.2 Modulation Transfer Function (MTF) analysis
The image sharpness was addressed through the analysis of its
Modulation Transfer Function (MTF) that represents, in the
spatial frequency domain and for a given direction, the image
spatial resolution. The transfer function of the system can be
obtained considering the response (Edge Spread Function -
ESF) of the optical system to an ideal edge (rectangular pulse).
Multiple methods have been proposed for determining the MTF
of remote sensing systems in-orbit. Most of the procedures use
specific artificial or natural targets on the ground for estimating
the Edge Spread Function (ESF). Through the first derivative of
the ESF it is possible to obtain the Line Spread Function (LSF),
whose Fourier Transform provides the Modulation Transfer
Function (MTF) (Figure 3).
Bright Side
Figure 3. Edge MTF estimation method.