32 
H 
33 
Phases | and 2 consist of two consecutive conyolutions between 
the thinned matrix and a 3x3 window of value '1' elements. Each 
convolution: consists practically in'summing to each pixel of the 
matrix the values of the 8 adjacent pixels. 
  
data input 
thinned matrix 
M 
  
  
enhancement of the node areas 
(1st convolution) 
M 
enhancement of the node points 
Phase 2 within node areas 
(2nd convolution) 
| 
Y 
Phase 1 
  
  
  
  
  
  
  
| ordered acquisition of the coordinates 
Phase 3 of the node points (following the hori- 
: zontal or vertical lines) 
  
  
  
37 
Storage of coordinates 
on magnetic tape 
  
  
  
  
Figure 2 
The first convolution enhances the node areas as areas constitut- 
ed by pixels of value 3. 
The second convolution enables us to identify, within each node 
area, the node point as the relative maximum point. 
Then we mark in the original matrix the node pixels by.a value 
'n' different, from '0!' and '1'. In the third phasc, we proceed 
to the identification of the subsequent horizontal lines and to 
the consequent ordered acquisition of the coordinates of the 
node.points in the lines. 
For every line, the matrix is scanned to identify the leftest 
element (initial element). The line is followed, analysing for 
each of its elements encountered during the course an appro- 
priate right neighbourhood, as shown in Figure 3. The analysis 
of the^positions of this neighbourhood is made in the numeric 
order indicated in the second column of the figure.