The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008
1198
relative adimensional global error in synthesis (ERGAS-Erreur
Relative Globale Adimensionnelle de Synthèse) (Wald, 2002).
Fusion methods are used for gathering information from SAR
and Optical data. A comparative analysis described using L
band Jersl SAR and Landsat TM data. IHS, Wavelet, PCA and
High pass filtering (HPF) techniques are applied and mean,
standard deviation, coefficient correlation and entropy factors
are computed to qualify the fused images (Rokhmatuloh et
al,2003, Shi,2005). Pal et al. 2007 fused four bands 1RS-1C to
ERS-2 using PCA technique for enhancement of gathering
geological information.
In our study the quality assessment analyses were evaluated for
multi sensor data. An HH polarized L band PALSAR, an HH
polarized C band RADARSAT-1 image and a three band
SPOT-2 XS were used for image fusion. IHS, PCA, HPF,
DCW techniques were used for the fusion of each of
RADARSAT-SPOT and PALSAR-SPOT images. All fused
images compared visually and statistically to SPOT XS. Bias,
CC, DIV, SDD and UIQI statistical analysis derived from fused
images.
image were used as SAR data. A SPOT HRV-2 XS
multispectral imagery having three spectral bands with a 20 m
resolution were used as optical data. PALSAR (Phased Array
type L-band Synthetic Aperture Radar) data of ALOS satellite
(Advanced Land Observing Satellite) is a new SAR mission
which has observation capability of high spatial resolution. In
this study a fine single polarized (FBS) mode PALSAR imagery
with pixel size 6.25 m x 6.25 m and a Fine Beam 1 mode of
RADARSAT-1 image with pixel size 6.25 m x 6.25 m were
used.
RADARSAT1
PALSAR
SPOT-2
Date
28/05/2006
10/06/2006
14 /05/2006
Sensor
SAR Fine 1
PALSAR/FBS
HRV/HRG
Pixel Spacing
6.25 m.
6.25 m.
20m
Orbit
55139
2010
Flight direction
Ascending
Ascending
Processing
SGF
LI.5
Polarization
H/H
H/H
Swath
50 km
80 km
60 km
Incidence angle
37-40
41.5
L29.6
2. STUDY AREA
In this research, Menemen (Izmir) Plain on the west of Gediz
Basin in the Aegean Region of Turkey was selected as a study
area that covers about 400 square km. The Aegean Sea lies on
the west of the study area, and Manisa Province lies on the
North (Figure 1). The study area covers both residential and
agricultural areas. The area has a micro relief however the slope
in general is 1 %. Texture and slope are important soil
characteristics for SAR backscattering. 3
Figure 1. Memenen plain
Table l.Data information
Before the image fusion process SAR images were pre-
processed by the commonly used speckle reducing filter
techniques. For the filtering of SAR images different sized
kernel windows were used with Gamma filtering and 3x3 one
was chosen.
In this study image fusion was conducted at the pixel level. In
order to avoid the combination of unrelated data spatial
registration accuracies should be at the sub pixel. Therefore in
fusion applications geometric correction is very important for
registration of the images. After reducing the speckle effects of
SAR images, SAR images were registered to SPOT image by
using image to image rectification method with a root mean
square error of less than 1 pixel. Cadastral maps in 1/5000 scale
and topographic maps in 1/25000 scale were used for the
rectification of SPOT images.
In this study 4 fusion methods have been examined namely
Intensity, Hue and Saturation (IHS), Principle Component
Analysis (PCA), Discrete Wavelet Transformation (DWT) and
High Pass Filter (HPF).
3.2 Pre-processing
3.3 Image Fusion Methods
3. MATERIALS AND METHODOLOGY
3.1 Data
Two different SAR data from different sensors namely
RADARSAT-1 and PALSAR were both fused with SPOT-2
dhta (Table 1). Although the PALSAR and the RADARSAT-1
images have the same resolutions and polarisations, images
were gathered in different frequencies (L band and C band
respectively). Particularly in this case the operating frequency is
a key factor in the penetration depth to see the affects on
extracting information from fused images. An HH polarized L
band PALSAR and an HH polarized C band RADARSAT-1
IHS method separates the Intensity, Hue and Saturation
components of a RGB image. Three bands of an MS image are
converted from RGB colour to IHS colour. Spatial frequency
related I component is replacing the high resolution image and
back transformation IHS to RGB is required (Pohl and Van
Genderen,1998).
PCA converts a multivariate data set of inter-correlated
variables into new uncorrelated linear combinations of the
original values (Pohl and Van Genderen,1998, Teggi et al,2003).
The difference between IHS and PCA is that while IHS is used
for 3 bands, PCA method can be used for more than three bands.