396 
Figure 2. t-ratio segmentation of forest stands in a 5x5 km area. Upper left: input composite SPOT XS,PAN 
bands, pixel size 10 meters. Upper right: information on landuse, hydrology and administrative boundaries, 
digitized from public maps at scale of 1:10 000. Lower left: initial regions used (Directed trees 
algorithm). lower right: result of t-ratio segmentation. Average region size: 2.4 hectares. Copyright: SPOT 
image, CUES. 
Algorithm 1. t-ratio region growing. 
Parameters: 
Hires t-ratio threshold. 
Iter Number of iterations. 
Max, Min Maximum and minimum 
region size. 
Functions: 
Nearest(R) Returns the ID of the 
adjacent region 
closest in feature 
space to region R. 
Returns the t-ratio 
for two regions R,N. 
Returns the size of 
region R. 
Returns the total size of 
regions R + N. 
START 
Index = 0 
WHILE Index < Iter DO 
Index = Index + 1 
T = Hires * (Index/Iter) 0,5 
REPEAT 
FOR every region (R) DO 
N = Nearest (R) 
IF (t-ratio(R,N)<T) and (size(R,N)<=Max) THEN 
List R and N for merging 
Merge all pairs of regions listed 
UNTIL no more regions are merged 
END WHIIE 
REPEAT 
FOR every region (R) DO 
IF size(R) < Min THEN 
N = Nearest(R) 
List R and N for merging. 
Merge all regions listed 
UNTIL size of every region >= Min 
STOP 
t-ratio(R,N) 
Size(R) 
Size(R,N)