which could automatically vary the integrating windows in 
according with the features in the sub-image, is presented to 
detect whether the noise dot or the target. The maximum 
window is 9x9. 
4) The wavelet coefficient of the optical image can be regarded 
as the fusion value when the wavelet coefficient of the SAR 
image may be noise dot. Otherwise, the fusion value is 
calculated using equation (4) in the window which the noise 
dot is detected. 
5) Reconstructing the SAR image using lifting wavelet 
technique. 
The scheme is followed as: 
Figure. 1 The scheme of the image fusion 
5. THE EXPERIMENT AND CONCLUSION 
Fig. 2(a) is a Radarsat SAR original image. Fig. 2(b) is a TM 
original image. The intention of the image fusion is to integrate 
the information of the TM image to SAR image. The 
Daubechies9/7 wavelet is used to decompose the images with 
the lifting means, and level is 3. The first SAR sub-image is 
handled with regional average method. Then the SAR and 
optical sun-images are integrated with the power average, and 
the powers are respectively 0.6 and 0.4. When integrating the 
high-frequency sub-images, the ratio method of the variances is 
used after the noise dots have detected along the horizontal 
direction and vertical direction using the ratio method of the 
energy. The experiments could be present Fig.2(c): 
The intention in the paper is that the fusion image can be used 
to extract the features well in order to match the real-time 
image (real aperture radar image) and refer image (fusion 
image). As a result, we respectively extract the features using 
the SAR original image and the fusion image [8] . Fig.3 shows 
that the fusion image could detect the targets well. 
Fig. 3(a) and (b) are the results of the SAR image and the 
fusion image using the LOG features extraction. Fig. 3(c) and 
(d) are the results of the SAR image and the fusion image using 
the FACET features extraction. A mass of the noise dots have 
been retained in the Fig. 3(b) and (d), and the continuity of the 
features is more obvious in the Fig. 3(d). The experiment 
proves that the fusion image is better than the original SAR 
image for feature extraction. 
(b) TM image 
Figure.2 The original image and fusion image 
(a) SAR image 
(c) Fusion image