1144
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008
image
band
mean
standard
deviation
entropy
Correlation
coefficient
with XS
image
average
gradient
Correlation
coefficient
with Pan
image
XS image
5
75.183
21.675
6.0952
1
4.0176
0.6541
4
71.941
15.599
6.0627
1
2.8171
0.7284
3
66.227
18.682
6.0598
1
3.2773
0.1661
HPF fused
5
75.113
22.141
6.3509
0.9455
6.6193
0.7214
image
4
71.871
15.884
6.2881
0.9636
6.5239
0.7956
3
66.157
21.041
6.2971
0.9843
5.6477
0.2985
SFIM fused
5
75.161
22.966
6.291
0.9548
8.0911
0.7365
image
4
71.810
17.516
6.2199
0.9603
7.3351
0.8139
3
65.971
19.383
6.2274
0.9537
7.2359
0.3127
ML fused
5
100.228
23.161
6.2391
0.9277
8.1029
0.8883
image
4
98.153
19.267
6.2595
0.9433
7.6337
0.9140
3
93.150
19.167
6.0762
0.8425
7.952
0.6674
Brovery
5
47.925
15.982
5.6371
0.7847
9.9214
0.9543
fused image
4
47.965
21.041
6.2393
0.9732
9.1894
0.6699
3
40.908
9.767
5.1429
0.7844
9.266
0.7208
IHS fused
5
73.574
14.842
5.9055
0.8231
6.5412
0.8970
image
4
71.761
22.728
6.4812
0.8786
6.8252
0.7000
3
64.028
21.48
5.8278
0.8157
6.5109
0.4894
PCA fused
5
136.353
28.614
6.9469
0.6441
10.3735
0.9613
image
4
134.02
13.613
6.9428
0.9546
9.396
0.7738
3
137.837
17.966
5.9238
0.2372
9.6198
0.9734
Table 3. Table of evaluate parameters
4.1 Parameters Statistics of Fused Image
The original multi-spectral images using XS to replace, and
panchromatic images with PAN replaced, evaluate parameters
are shown in the table3:
From the parameters of table 3, we can see that:
(1) All fusion method in accordance with the definition in
descending order, the order is: PCA>Brovery>ML>
SFIM>MIHS>IHS>HPF;
(2) All fusion method in accordance with the Spectra maintains
degrees in descending order, the order is: HPF>SFIM>
Brovery>MIHS>ML>IHS>PCA; 3
(3) All fusion method in accordance with the entropy in
descending order, the order is: PCA>MIHS>HPF>IHS>SFIM>
ML>Brovery.
4.2 Feature Identification Accuracy of Fused Image
Different fusion methods have different impacts on image.
Image Recognition is the application of spectral characteristics
and structural characteristics of different features to identify
information; therefore spectra and texture information on the
objectives of the interpretation are important significance [10] .
In order to verify the influence of various fusion methods on the
classification accuracy, in this paper, the image data of different
experiments using the same processes to deal with unsupervised
classification; and make classification accuracy
test, select high precision fused image to make supervised
classification.
4.2.1 Research Methods: Make classification with maximum
likelihood classification; using random method to select 256
ground inspection points, make accuracy test for thematic maps
of XS image and fused image, obtain total accuracy and Kappa
index.
4.2.2 Accuracy Test of Unsupervised Classification
From the comparative data table 4, we find that: PCA fusion
image, in addition to other fusion image classification accuracy
are significantly higher than those without fused image
classification accuracy, the reason maybe that: PCA fused
image has the worst spectrum distortion, and it leads to the
lower classification accuracy. Descending order of the
classification accuracy is: SFIM> HPF >ML>Brovery >
XS>IHS>PCA.
type
XS
image
HPF fused
image
ML fused
image
Brovery fused
image
PCA fused
image
SFIM fused
image
IHS fused
image
Total
77.34%
81.25%
80.47%
78.52%
67.97%
84.38%
76.95%
accuracy
Kappa index
0.6799
0.7468
0.7298
0.6809
0.5271
0.7810
0.6454
Table 4. Comparative data of image unsupervised classification accuracy