26 
Pixel- 
based 
No Grouping 
Choose max 
Window- 
based 
Single-scale 
GrouDins 
Weighted 
Average 
Region- 
based 
Multi-scale 
Grouping 
AMO 
Results 
Schemes 
Activity Level Grouping method Combining method 
Figure 2. The flowchart of the image fusion procedure 
Consistency 
Verification 
Table 1 Some used abbreviations in the paper 
Abbreviation 
Description 
AMO 
Adaptive multi-objective optimization 
CBA 
Coefficient-based activity level 
CM 
Choose-max coefficient combining 
DWT 
Discrete wavelet transform 
LPT 
Laplacian pyramid transform 
MG 
Multiscale grouping 
MORPH 
Morphological pyramid 
NG 
No grouping 
RBV 
Region-based consistency verification 
SiDWT 
the shift invariant DWT 
WA 
Weighted average coefficient combining 
WBA 
Window-based activity measurement 
ADD 
LPT MORPH 
(a) WA-WBA+NG+CM+NV 
DWT 
MORPH DWT SiDWT 
(b) WA-WBA+NG+AMO+RBV 
SiDWT 
Figure 3. Results of two fusion methods 
Table 2. Performance of Fig.3 (b) method when Using Different decomposition levels 
Combination 
MSD(Levell) 
SD 
EN 
CE 
MI 
UI 
WA-WBA+ 
NG+ 
AMO+ 
RBV 
LPT 
61.7211 
7.3811 
2.1661 
3.2951 
0.7036 
Morph 
61.5777 
6.7915 
2.6465 
5.7916 
0.7212 
DWT 
61.1049 
7.5614 
1.9310 
2.7441 
0.7090 
SiDWT 
61.5212 
7.1839 
2.4275 
5.1667 
0.7231 
Trough the 
Morph, DWT, 
can find use tl 
combing coeffi 
better quality tl 
1), we find w 
region easily i 
speckle noise, 
space resolutioi 
in the image 1 
water region).", 
origins (e.g. 
resolution of t 
needed. Throu 
distinguish wa 
method (b) tha 
multi-objective 
combing to sea 
3(b). Therefore 
search the ada 
can save up th 
huge different 
simple combii 
performances. 
Meanwhile, 
images from d 
table 2, the MI 
than do the 01 
MORPH the 
measure indica 
For poor fusio 
MI assigns the 
the DWT fusic 
lower than Si 
observers. CE 
remarkable in 1 
As one can < 
SD, to extract 
apparent corre 
evaluations. El 
in the fused im 
values for the 
sensitivity of 
fluctuations in 
discriminate b<