International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B1. Istanbul 2004
2.3 Component Development for Fusion
Main components to be developed in this study were data
input component, filtering component for image improvement,
image analysis component for such as geometric correction of
images, image fusion component and output component. For
image fusion component, modules for IHS and Wavelet method
were developed.
3. EXPERIMENT BY FUSION METHOD
3.1 IHS Fusion Method
The image fusion method using color models of IHS(Intensity-
Hue-Saturation) had been firstly used by Hydan in 1982. Since
then, it has been the most common method. IHS fusion method
applies IHS transformation to three low resolution images
existing on RGB color models and creates intensity images,
hue images and saturation images(Jensen, 1996).
IR HC)
>
es ra RCE
er | 1)
LqE-G)GO - EJ]
H=tos | s
VE GEXU-BIG-E|
Whereas, intensity images(grey image) include information
related to spatial resolutions of low resolution images and hue
images and saturation images contain data related to spectral
resolutions. Thus, when intensity images are replaced by high
resolution images with good spatial resolutions, spectral
resolutions of low resolution images and spatial resolutions of
high resolution images are combined. These images become
fusion images of high resolutions containing plenty of spectral
resolutions of low resolutions through RGB transformation.
IHS transformation method is easy to apply, but allows only
three bands to be applied.
3.2 Wavelet Fusion Method
Multiple resolution Wavelet transformation has been recently
studied. It has been applied to signal processing and image
processing in various ways. In particular, it has been widely
used for image compression, boundary extraction, object
recognition and image fusion in image processing (H.H.Kang
et al., 2001; Garguet-Duport et al. 1996).
When applying Wavelet transformation to images, one rough
image and three precise images are obtained. At this point, a
rough image contains data related to spectral resolutions of
images and three precise images contains spatial resolutions
according to directivity. Therefore, Wavelet fusion method
applies Wavelet transformation to high and low resolution
images using properties of Wevelet transformation, replace a
rough image of low resolution image by that of high resolution
image and then, carries out Wavelet reverse transformation
using a rough image of low resolution image and precise
images of high resolution images. Wavelet transformation
formula is defined as follows(Sunar et al. 1998).
Wa, b= [ A) C0), a»0, be R
v la] a (2)
a : scalar Factor
b : translation Factor
(1) : mother wavelet
Approximation
Image (change)
Horizontal detail
image
Vertical detail
image
Diagonal detail
Figure 2 1 pass diagram of Wavelet method
Figure 2 describes image fusion processes using sample
images applied in this study. Firstly, as applying two-step
Wavelet transformation to KOMPSAT with high spatial
resolution, approximate images and detailed images with the
same size as LANDSAT images with low spatial resolutions are
obtained. Next, approximate images on the second step are
replaced by each multi-spectral image and then are reversely
transformed. Accordingly, the images are fused as multi-
spectral images with three spectral bands having high spatial
resolutions.
4. DEVELOPMENT OF IMAGE FUSION COMPONANT
This study created component-based program using object-
oriented concept to perform pre-processing function and image
fusion function for image fusion. Figure 3 shows overall layout
of image fusion system. Figure 4 explains links among classes
— CManage
Source Image
IMAGE Result |
Low resolution Fc esuit Image
y Management
High resolution »
e d»
CFusion ——— A)
IMAGE —— ei uc
Fused high e Wavelet
Resolution x Fusion Method
lee ul Class Selection
Figure3 Diagram of Fusion System.
and data flow for image fusion. System implementation largely
consists of image input sector, fusion sector and output sector,
in which fusion sector contains fusion modules and image filter
modules of IHS and Wavelet.
Source & Result
Image FR Method Fusion
Management election Execution
3 a
HR
[ems ]!
tH
| CManage | CIToR
[mses |
{
| CFusion | >
—
uit
Figure 4 Work Flow of Class and Data
Intern
Rawr
and C
Each
comp:
4.1 D
Co
into t
Classi
analy:
explai
4.2 Di
Progr:
windc
File, |
Subm
Rougl
S.1AJ
This
aroun
from
Marcl
select
Geom
define
rearra