The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008
1133
2. Entropy:
The entropy of the image is an important index to weigh the
information capacity, according to the Shannon’s
information theory principle, the definition of entropy is:
where, M(i,j) , F(i,j) are the pixel value of original
image and fused image in certain component, respectively.
Correspondingly, M(i,j) and F(i,j) are the means. M
and N are the weight and height of the image.
4. Average Gradient (A. G.)
(8) Average gradient is able to accurately reflect the fine
contrast of the image. Generally, the larger the average
gradient is, the clearer the image is.
where, P is the ratio of the i pixel number to the total
pixel number in the image, m is the number of grades.
Since the entropy represents the information capacity, the
larger the entropy is, the more the image contains
information. Seen from formula (8), we know that the
maximum value of H occurs if and only if each P is equal.
It means that the information capacity tends to be the
maximum if each gray grade has almost the same account
pixels. By comparison of the entropies of different images,
we can judge the ability of expressing detail information.
3. Correlation Coefficient (C. C.)
The correlation coefficient between the original image and
the fused image shows the similarity in small size structures
between the original and the fused images. It should be as
close as possible to 1 (Luclen, 1997). The correlation
coefficient p is:
VG = ^J]yA/ 2 (x,y) + A)/ 2 (x,y) (10)
MV Jm ,
where, Axf(x,y) and Ayf (jc, y) are the difference of
f(x,y) at x and y direction in certain component,
respectively.
5. Average Spectral Distortion (A. D.)
The definition of the spectral distortion in the Mi
component is:
p =
Z Z (Af a, j) - M(i, j))(F(i, j) - F{i, j))
I M /V IYJ
III (■m,j)-M{i,j))' TX (F(i, j) - F(i, j))
(9)
where, V* and V' k are, respectively, the gray values of
multispectral image and fused image in the Mi component
at position (ij). As the spectral effect is expressed by the
synthesized of the multiple bands, we employed the
average spectral distortion, noted as D, to illustrate it. The
greater the D is, the larger the distortion is.
C.C with
Image
band
Mean
S.D.
Entropy
multi
spectral
SAR
A.G.
A. D.
Multispec
tral
1
89.3915
2699.9938
7.2407
1.0000
—
16.1554
—
2
100.7079
2370.8928
7.2342
1.0000
—
16.1896
—
3
76.6344
2438.7121
7.0459
1.0000
—
15.9443
—
SAR
-
57.7986
2438.6782
7.0837
—
1.0000
11.8878
—
1
36.2861
3745.5817
6.0136
0.1195
0.7344
20.4025
Brovey
2
43.7587
4371.4152
6.3431
0.1315
0.6349
28.0559
212.7874
3
39.4694
4757.6124
5.9325
0.1931
0.5908
29.3560
Triangle
IHS
1
55.3576
2411.2494
7.0468
0.1136
0.9457
13.4192
2
66.3127
2635.3794
7.3110
-0.0064
0.9334
15.6292
165.2970
3
47.5083
2397.7407
6.7985
0.1503
0.8508
14.3046
1
30.6315
106.7828
4.8998
0.0962
0.6671
2.0525
PCA
2
29.9306
175.8581
5.5172
0.2041
0.5935
2.1249
180.1813
3
29.6898
110.6867
4.9486
0.1016
0.6612
2.1421
Gram-
Schmidt
1
20.7771
172.5953
5.4618
0.1937
0.8778
3.0947
2
40.2084
361.7482
6.2062
0.1502
0.8136
3.8246
195.1719
3
10.9059
146.2115
4.8158
0.1034
0.9003
2.8324
Normal
1
40.0525
482.7768
6.3879
0.9572
0.2869
6.7017
wavelet
2
54.6524
649.9612
6.6190
0.9684
0.2852
8.3276
127.3534
transform
3
42.6195
663.5351
6.5742
0.9690
0.2548
8.2042
The novel
1
83.3250
1859.9954
7.3748
0.8014
0.3683
12.1231
65.1349
method
2
92.1778
1652.6537
7.3236
0.7828
0.4196
11.7862