l-pixels and a connected component of 0-pixels 
which surrounds it directly. A hole border is 
defined as the set of l-pixels located between a 
connected component of 0-pixels and a connected 
component of l-pixels that surrounds it directly. 
For both outer borders and hole borders, the bor- 
der itself is defined by a set of l-pixels. The 
O-pixels are never used to define a border, ex- 
cept in the case of the frame of the image that is 
considered a special hole border. For any given 
connected component of l-pixels there is one, and 
only one, outer border and it is unique. There 
are two algorithms that were implemented for 
border following. The theory behind both of 
these algorithms was developed by Suzuki and Abe 
(Suzuki and Abe, 1985) and the details are given 
there. 
Generate Freeman Chain Codes 
  
Once the border pixels have been found, the 
binary image must be converted into a format that 
can be used to form line segments. The Freeman 
chain code is used to represent a boundary with a 
sequence of numbers, each of which indicates the 
change in direction from one border pixel to the 
next. An 8-direction chain was used. Each num- 
ber corresponds to a particular change in direc- 
tion from one pixel to the next. For example, the 
number 0 represents a change in direction of 0 
degree from one pixel to the next. The angular 
change is measured from the horizontal to a line 
formed by joining the current pixel in the chain 
to the next pixel in the chain. The number 4 
represents a change of 180 degrees from one pixel 
to the next, etc. One entire chain code is repre- 
sented by a list whose first two elements are the 
(i,j) coordinates of the initial point in the 
chain, followed by a sequence of numbers each of 
which comes from the set {0, 1, 2, 3, 4, 5, 6, 
7}. Each of these numbers represents the change 
in direction from one pixel to the next. The 
entire binary image will consist of many such 
chains. To form the chain codes, a tracking 
routine was developed that examines the nearest 
neighbors of a pixel in the chain to find the 
next pixel in the chain. An index array, which is 
as big as the original image and is initially set 
to contain all zeros, is used to indicate if a 1 
has already been included in a chain or not. The 
input binary image is scanned by starting with 
the pixel in the upper left-hand corner of the 
image. When a 1 is found at the location (i,j) 
and the contents of the index array at (i,j) is 
0, a new chain is started. As each pixel in the 
chain is found, the appropriate number from the 
set f0, 1, 2, 3, 4, 5, 6, 7} is placed into the 
chain list and a 1 is placed into the correspond- 
ing location of the index array. When the scan 
reaches the pixel in the lower right-hand corner, 
the algorithm terminates. 
Polygon Approximation 
An arbitrary, two-dimensional, open or closed 
digital curve is represented by a Freeman chain 
code and consists of an ordered set C of N pixels. 
The purpose of polygon approximation is to replace 
each digital curve with a consecutive series of 
line segments, each of which are within a certain 
error of the original curve. There are many poly- 
gon approximation techniques, and the theory 
behind the one used in this research was developed 
by Ramer (Ramer, 1972). The pixels in the set C 
can be considered as the vertices P; of a polygon 
if the curve is closed. If the curve is not 
811 
closed, the set C can be considered as the verti- 
ces of a consecutive series of line segments. In 
either case, the task of the polygon approxima- 
tion algorithm is to provide a reduced set C'with 
a smaller number of edges N'and whose vertices B. 
coincide with those of the original set C. 
RESULTS 
The result of applying the above mentioned algo- 
rithms to the radar image shown in Figure 1 is 
given below in Figure 3. The region properties 
of area, orientation of the axis of least inertia, 
and the measure of region spread were the only 
properties required to obtain this final result. 
It can be seen that the runway patterns have 
clearly been delineated with line segments and 
the other components in the image have been elim- 
inated. The image is virtually noise free and 
presents a good extraction of the runway pattern 
shown in Figure 1. 
  
Figure 3. 
Polygon Approximation. 
CONCLUSIONS 
l. The approach taken to extract airfields from 
radar imagery is valid. 
2. Although the method of edge reinforcement 
using relaxation is computationally intensive, it 
produces excellent results. 
3. The components of the airfield were extracted 
using the three region property calculations of 
area, orientation of the axis of least inertia, 
and the measure of region spread. 
REFERENCES 
+ References from JOURNALS: 
a) Nevatia R. and Babu K.R., 1980. Linear 
Feature Extraction and Description. Computer 
Graphics and Image Processing, Vol. 13, PP. 
257-269. 
b) Ramer, U., 1972. An Iterative Procedure for 
Polygonal Approximation of Plane Curves.  Com- 
puter Graphics and Image Processing, No. 1, 
pp. 244-256. 
c) Schachter, B.J., Lev, A., Zucker, S.W., and 
Rosenfeld, A., 1977. An Application of