International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004 
3. IMAGE MATCHING AND MANAGEMENT 
[n this section, image matching method and satellite image 
management will be discussed. Figure 2 shows the structure 
diagram for realizing them. The system consists of two major 
components, compression and matching part. 
  
GCP images 
  
Search Region 
  
  
  
  
  
  
          
  
  
  
  
  
  
  
     
  
  
  
  
  
  
X 
À Y 
\ 
> WT 4 
Y 
Parameters al Matching 
Extraction > 
  
  
  
— 
  
D 
x 
uorsso1duio 
  
  
  
Figure 2. The whole structure of proposed system 
The raw images obtained from satellites or off-line medium are 
stored with bit stream after compression. These images will be 
decompressed when they need to be corrected geometrically. In 
case of GCP image, after all GCP images are applied to wavelet 
transform, some parameters which can express properties of the 
GCPs are stored. These parameters will be compared with that 
of raw image for matching. 
1.0 Satellite Image Compression 
Image compression is an important work in image processing. 
The common feature for most of application using image is that 
without compression, the data to be handled is too large. For 
example, in order to store one scene of Landat-7, storage over 
300MB is required. In this section, the efficient compression 
method which was introduced to (Park, J.H., 2002) is briefly 
referred. 
As stated (Park, J.H., 2002), the proposed method is based on 
the fact that some region in spatial image and wavelet 
coefficients in time-frequency domain are closely related. The 
method could be implemented by applying proper coding 
methods to each block which are non-overlapped region of 
spatial image. 
The region or block with low activity in the spatial image, so- 
called background region(BR), appears not only as insignificant 
coefficients in the wavelet domain, but also has little influence 
in the reconstructed image. In general, compression efficiency 
for these regions is known to higher than complex region. On 
the other hand, high activity blocks, edged region(ER), are 
related to significant coefficients, which give a great influence 
to image reconstruction. Observably, the high activity regions 
are mostly linked to edges or boundaries of images. 
The objective of our proposed image compression algorithm is 
to obtain an efficient coding result by using the relationship. In 
order to accomplish our purpose, BR regions are encoded by 
color information coding method, which is similar to encode the 
inner parts of the object in the conventional object oriented 
coding(OOC); this scheme is appropriate to represent the 
region with low activity. To encode ER regions, a technique 
similar to the conventional bit plane coding is applied to ER 
region; this scheme is efficient to represent significant 
coefficients by connecting to quantization stage. Moreover, this 
method is feasible to implement a progressive transmission by 
simple operation. More detailed information about this method 
can be obtained from [1]. 
Compressed data stream is archived in the RAW database as 
shown at figure 1. 
2.0 Image Matching 
Experiment for image matching using Wavelet 
In order to investigate that image matching using wavelet is 
possible, the simple experiment is performed. Figure 3-(a) is the 
original image and Figure 3-(b) is the changed image that is 
increased both contrast and brightness about 20% compared to 
the original image. 
  
342 
(a) Original image (b) Changed image 
Figure 3. Experimental images for analyzing relation between 
image matching and wavelet 
We applied wavelet transform to above two images, and 
investigated energy of each image in transform domain. The 
result is shown at table 1. 
  
  
  
  
  
  
  
  
  
  
  
(A) (B) (C) 
Original Brightness Brightness(+20) 
= (+20) Contrast(+20) 
3 HL 80.5 80.5 125.5 
lov LH 136.0 136.0 211.0 
HH 20.9 20.9 32.6 
5 HL 85.0 85.0 132.2 
Diver LH 75.1 75.1 116.1 
: HH 19.3 19.3 30.1 
HL 51.9 51.9 80.8 
Layer LH 37.8 37.8 58.8 
2 HH 6.1 6.1 9.6 
  
  
  
  
  
  
  
Table 1. Energy of wavelet domain 
The energy is unchanged when brightness is only changed from 
the table. However, energy is increased when contrast is 
changed and brightness as well, but degree of the increase is the 
same to that of original. For example, in case of original image, 
the energy of HL in layer 3 is corresponding to 15.7% of total 
energy and column (C) is equivalent to that. 
When the satellite image in the DAW DB needs to be corrected 
geometrically, the image matching process will be operated 
after the image is decompressed. The image matching in our 
Interna 
system 
satellit 
Figure 
  
  
In the 
The s 
the la 
non-o 
TE js 
lowest 
The n 
repres 
howe 
differ: 
directi 
0, ver