ori ginal 
image 
  
  
  
  
Quadtree generation based on spectral 
and statistical information 
  
  
whi dn 
Segmentation 
quadtree 
  
  
3 
Low — level 
quadtree generation 
  
  
  
Geometry and attribute transformation, 
construct homegeneous region list 
  
  
  
  
  
Homogeneous 
region 
quadtree generation 
High — level 
quadtree 
  
  
  
  
  
  
Control and feature 
quadtree generation 
  
  
   
   
  
   
     
  
Spatial 
processes till 
  
  
  
  
   
relation 
quadtree 
Repeat above 
based on 
output 
satisfaction 
DGs 
  
    
Figure 1. Hierarchical processing for spatial relation extraction based on quadtree. 
2. A NEW PROCEDURE FOR SPATIAL 
RELATION EXTRACTION BASED ON 
QUADTREE REPRESENTATION 
The steps of automatic spatial relation extraction based 
on quadtree representation by a computer system are 
as the following and Figurel shows the hierarchy of 
the processing in which; 
Stepl. The image is first segmented by using quadtree 
generation . The characteristics are detected by only 
checking the spectral and statistical attributes, and a 
segmentation quadtree is made. 
Step2. A segmentation quadtree is defined as a 
binding which composed of location code for a 
558 
quadtree node, attribute values measured during the 
scanning of a quadtree generation and region 
number. Then the quadtree segments from step 1 are 
merged to yield a new quadtree segment by defining 
the segment descriptive parameters and mergence 
criterion that relate to the thematic units. At the same 
time, the spatial configuration characteristics such as 
shape , size etc. are extracted. Througe step 2 , the 
segmentation quadtree is transformed into the 
homegeneous region quadtree. 
Step 3. The spatial relations and constraints among 
the segmentation quadtree are extracted by using 
control and feature quadtree generation. Thus the 
region relationship list is formed. As more and more 
constraints are filled in the list ,each spatial relation 
[eS