on de 
a mise 
stemes 
ise de 
-BULL. 
Riz. 
stein, 
Image 
is" on 
SMC-3, 
1 And 
Inter 
f on 
ources 
9-431. 
on Of 
Boston 
ories, 
n et 
ations 
cional 
ise en 
Image 
que de 
de 
ce des 
: de 
Paris- 
re des 
ese de 
ouse. 
undary 
ol. C- 
1, ÀA., 
exture 
. IEEE 
and 
1976, 
SPATIAL RELATION EXTRACTION FROM REMOTELY 
SENSED IMAGE ON QUADTREE REPRESENTATION 
Li Deren 
Guan Zequn 
Department of photogrammetry &. Remote Sensing 
Wuhan Technical University of Surveying and Mapping 
Wuhan 
ABSTRACT 
This paper addresses a development of spatial relation 
extraction from remotely sensed image based on 
quadtree representation. The input image is first 
transformed into a hierarchical structure of quadtrees, 
which is composed of quadtree, segmentation 
quadtree; homogeneous region quadtree and spatial 
relation quadtree. With the description of hierarchical 
quadtrees ,appropriate spatial relation can be extracted 
‚which is from local to global as more and more object 
knowledge is used. The strategy that is called 
discrimination graphs algorithm for spatial relation 
extraction in terms of quadtree representation is 
presented. . The result obtained in this paper is clear 
and can easily be used for remotely sensed image 
interpretation. 
KEY WORDS; 
Quadtree representation, Hierarchical quadtree, Object 
Spatial relation extraction , 
knowledge, Discrimination graphs. 
1. INTRODUCTION 
The region quadtree has been applied in both raster 
image and raster map in various forms. Recent 
advance in the use of quadtrees for computer image 
processing and computer cartography have made 
efficient algorithms for conversion between the region 
quadtree and other image representation . In most 
published quadtree research, we has seen the space 
efficency of a quadtree depending on the particular 
structure used to represent it. However, we also saw 
that it lack close links with image interpretation. 
557 
China 
Most research in region quadtree so far was focussed 
on developing image encoding ,storage , transformation 
etc. (Jean, 1985).In this paper the term “ quadtree 
3 
representation we mean here some hierarchical 
structure, which is composed of quadtree, 
segmentation quadtree, homogeneous region quadtree 
and spatial relation quadtree ,generally concerning the 
local properties that stem from spectral and statistical 
information and the global properties that originate 
from prior object knowledge. 
In many situation ,it does not suffice to determine the 
mapping between region quadtree , which stemming 
from the low — level quadtree generation, and the 
associated hige — level quadtree representation. This 
indeterminate is not only in the meaning, but also in 
the spatial areas. It must result in indefinite relations 
between the high — level quadtree representation to 
convert the indeterminate region quadtree into the 
high — level quadtree representations. Thus ,when we 
interpret quadtree using relations in terms of spatial 
knowledge, the result shall still be indeterminate. In 
our work we have attempted to solve the problems by 
taking a strategy which has a feedback route to revise 
the errors under current best representation. As some 
processes are repeated, more attributes and relations 
are discovered in the quadtrees, which force each 
representation to become more specific. This approach 
is therefore referred to discrimination graphs (Jan, 
1988).