X k-up-table 
n by index 
assified or 
useful for 
oral image 
n and their 
ding noise 
arithmetic 
g different 
to 3 steps. 
a temporal 
p program 
arithmetic 
limited on 
expression 
ORTRAN 
on of the 
am checks 
roceeds to 
in order to 
alues 0 is 
Maximum 
following 
tation the 
Input files 
stics) file. 
SEAN 2.0 
  
can classify up to 255 classes, class O is always reserved for 
unknown (unclassified) class. 
  
  
Number of channels af the Input image : 4 
The formule can be up to 68 characters long 
The formula will be constructed based on FORTRAN style 
using any FORTRAN intrinsic functions and the following 
  
mathematics character set : 
1234567698.[ j-?f- x Mathematics symbols 
01,52,....bi,...bn ^^ Digital values of band í 
Wew value - FID, 2... ba] = 
|b3-bzi*t 00.4 b3:b2)4127. | 
  
  
  
  
  
  
  
  
Figure 3. Formula definition for Arithmetic operation 
Program first classifies the look up table LUT1 and then 
the image data in CPT format will be labeled according to the 
code of pixel value vector that has been classified in LUTI. 
During classification program prompts for two parameters : 
sigma range and correction factor. The sigma range is actually 
multiplicity of standard deviation of the class. Default value of 
it is set to 3.0. The correction factor is primary conversion 
coefficient. Decrease of it has the effect to falsely increase the 
likelihood for the class. Its default value is set to 1.0. When 
classification result seems to be incorrect by unknown reason, it 
is possible to use this coefficient to correct the result. The 
classification process is interactive that is after classification 
user can observe at once the result on the screen and if the result 
does not meet the requirement then by changing sigma range 
and correction factors classification can be repeated until 
satisfied result is achieved. 
Class 1 : POND 
Sigma Range — Correction Facter 
Ws; 
  
  
  
  
Figure 4. Maximum likelihood classification 
3.5 Change Analysis 
This is a group of three separate functions used to display 
multidate image after registration, extraction of the change 
between two classified images and display of the change matrix. 
   
By Plural image display menu two or three multidate 
images can be mixed and displayed. The nature of the mixed 
display is to assign certain band of input images to one of three 
image planes red, green or blue. The result image is an overlay 
of different bands from different images. WinASEAN can mix 
image data in different format as BSQ, BIL, BIP. 
In order to generate view of changes based on two 
classified images taken from different time ASEAN create at 
first change matrix and stores it in file .CHA. Program ignores 
all combination between the same classes and assigns them 
black color for display. The .CHA file has structure as following 
Image header 1024 bytes 
Encoded image data Number of lines * number of 
columns * code size 
Color table 
LUT2 Frequency of each change pattern 
Name of the first image 
Legend of the first image 
Name of the second image 
Legend of the second image 
LUTI Look - up - table containing all 
changes 
Because the change matrix is actually not image so in 
order to display it program assigns to each code three artificial 
gray values and then use it to generate visual presentation. The 
color is selected so that the color difference between codes is 
maximal. During display individual color presentation for each 
change combination can be changed. By this way some change 
combinations can be switched off by assigning to black color or 
enhanced some combination by changing their color to vivid 
one. 
1. mage 2. image Counis Code 
VEGET2 vegeti — 128920 12 
Figure 5. Change matrix display 
3.6 Bird's Eye View Image Generation 
The goal of bird's eye view image generation is to create a 
projected image according to the selected azimuth, depression 
angles and DEM of the selected area. DEM (Digital Elevation 
Model) contains ground height or elevation of each pixel and it 
is represented as grid with ground resolution as same as 
resolution of remote sensing data to be overlaid on the DEM. 
Each cell of the DEM contain information about elevation 
which is usually coded by 1, 2 or 4 bytes. 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B6. Vienna 1996