International Archives of Photogrammetry and Remote Sensing, Vol. 32, Part 7-4-3 W6, Valladolid, Spain, 3-4 June, 1999 
189 
inventory maps. The total length of the forest inventory stand 
borders was 30,380 m. 
Figure 4 shows the visibility result for the first interpreter: the 
total stand border length from inventory maps is set to 100% 
visible (best). In every image product more than 71% of the 
stand borders could be detected, whereby IHS_TM showed 
worst visibility. The visual part of the delineation increased in 
the following order: Pan, IHS_SP, Ortho and achieved 
maximum percentage at Qsim with 88%. 
Fig. 4. Visibility percentage of stand borders of interpreter 1. 
Figure 5 shows a comparison of the visibility for 3 image 
products and two interpreters. For both interpreters the visibility 
percentage is increasing from IHSJTM to Pan and Qsim. The 
class partial visible is decreasing with increasing geometric 
resolution of the image product. It is important to point out, that 
the fused IHS_TM achieves worse visibility than the single Pan 
image, i.e. in this case the fusion did not lead to enhanced 
information for visual analysis. 
Fig. 5. Comparison of stand border delineation of two 
interpreters with 3 image products. 
On the other hand, IHS_SP with the combination of IHS and 
AIF transformed SPOT data showed better results than the 
panchromatic image alone (Figure 4). 
3.2. Calculation of quality measure criteria 
The buffer method was used with 3 buffer widths (2.5m, 5m and 
10m). Only results with 10m will be presented here (10m is an 
acceptable delineation accuracy, as defined by the forest 
administration). Table 5 shows the result for correctness for 
interpreter 1: total delineation is the total length of all visually 
extracted stand borders and matched delineation is the total 
length of the stand borders inside the buffer around the 
reference. 
Interpreter 1 
PAN 
1HS_TM 
IHSJP 
ORTHO 
QSIM 
Matched Delineation 
19867 
17554 
20347 
22557 
23117 
Total Delineation 
23259 
22558 
23958 
25610 
26625 
Correctness 
0.85 
0.78 
0.85 
0.88 
0.87 
Table5. Calculated correctness results from interpreter 1. 
(delineation in m). 
The best results for correctness were achieved by interpreter 1 
with Ortho and Qsim with 88% and 89 % respectively, followed 
by IHS_SP and Pan with 85%. IHSJTM achieved the worst 
result with 78%. Interpreter 2 achieved the best results with 
Qsim with 78%, followed by IHSJTM with 74% and Pan with 
70%. So, the results of interpreter 2 were ca. 10% less accurate 
than the ones of interpreter 1. 
Table 6 shows the result for completeness for interpreter 1: total 
reference is the sum of all stand borders, and matched reference 
is the percentage of the reference network within the buffer 
around the delineated data. 
Interpreter 1 
PAN 
IHS_TM 
IHS_SP 
ORTHO 
QSIM 
Matched Reference 
21248 
21003 
23383 
23267 
23185 
Total Reference 
30380 
30380 
30380 
30380 
30380 
Completeness 
0.70 
0.69 
0.77 
0.77 
0.76 
Table 6. Calculated completeness results from interpreter 1 
(delineation in m). 
The best results were obtained with IHS_SP, Ortho and Qsim 
with 0.77 and 0.76. The worst result again with IHSJTM. 
Table 7 shows the result for quality for interpreter 1. Best results 
were achieved with Qsim and Ortho with 0.68 and 0.69 
respectively, IHS_SP is medium and again IHSJTM provided 
the worst quality.