The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008
1200
3.5 Statistical Inspections
In his study, an HH polarized L band PALSAR, an HH
polarized C band RADARSAT-1 image and a three band
SPOT-2 XS were used for image fusion. The quality assessment
analyses were evaluated for multi sensor data. In order to
evaluate the improved spectral quality of the image fusion
techniques, the original SPOT XS image was compared with
the 8 resulting fused images. IHS, PCA, HPF, DCW techniques
were used for the fusion of each RADARSAT-SPOT and
PALSAR-SPOT images. All fused images compared
statistically to SPOT XS using Bias, CC, DIV, SDD, UIQI
statistical analysis derived from fused images (Table 2 and
Table 3).
performance. On the contrary DWT gives the better results for
the RADARSAT-SPOT fused images.
Among the all fused images, statistical analyses show that IHS
technique gives the worst results for both RADARSAT and
PALSAR images (Figure 3).
RADARSAT-SPOT
Band 1 Band 2 Band 3
□ HPF □ PCA ■ IHS »DWT
■ tt>F»PCAalH5aCWT
Palsar
Bias**»
nrv %
SDD %
CO
UIQI
11PF B1
0.51
10.38
1.08
0.886
0,884
B2
0.72
9.88
1.66
0.902
0,900
B3
0.73
12.16
3.08
0.933
0,931
PCA B1
0.04
19.39
2.07
0.735
0.730
B2
0.12
28.81
4.76
0.611
0,603
B3
0.82
48.12
13.71
0.449
0,425
IHS B1
20.74
-524.63
-30.32
0.199
0,133
B2
28.93
-457.72
44.29
0.310
0,210
B3
22.19
-155.63
-29.35
0.644
0.561
DWT B1
0.11
19.17
2.04
0.777
0,772
U2
-0.88
27.87
4.94
0.761
0,750
B3
-9.13
59.92
17.98
0.616
0.554
Table 2.Statistical results for PALSAR fused images
Radarsat
Bias %
DIV %
SDD %
CC
UIQI
HPF B1
0.50
7.60
0.79
0.868
0,867
B2
0.70
6.94
1.16
0.884
0,883
B3
0.66
9.32
2.34
0.915
0,913
PCA B1
0.03
19.25
2.05
0.796
0,791
B2
0.09
21.95
3.82
0.714
0,708
B3
0.71
27.58
7.30
0.647
0,638
IHS B1
6.84
581.66
-32.57
0.308
0,205
B2
8.39
-522.62
-49.02
0.458
0,315
B3
1.84
-198.07
-35.61
0.774
0,671
DWT B1
-0.07
4.02
0.41
0.897
0,896
B2
0.12
2.31
0.38
0.915
0,914
B3
0.60
-2.10
-0.51
0.938
0,937
Table 3.Statistical results for RADARSAT fused images
Figure 3. Graphics of statistical analysis
A spectral analysis was done for agricultural areas using a
transect (Figure 4 and Figure 5). For the agricultural areas the
best results have been gained using HPF for the fusion of
PALSAR and SPOT XS (Figure 5 and Figure 6).
Figure 4. Transect on SPOT XS
Figure 5. Spectral analysis of SPOT XS
As seen from the tables HPF and DWT methods show better
results in both PALSAR fused and RADARSAT-1 fused
images. Statistical results show that the fusion results of
PALSAR-SPOT using the HPF method gives the best