14 
Fused image 
2.2 SAR texture extraction 
The Lee filter for SAR speckle removal is used to reduce the 
effect of the multiplicative noise on homogeneous areas at first. 
Thus, the ratio image M of despeckled SARcfe image to its 
lowpass approximation SAR L , obtained by the a trous wavelet 
algorithm as approximation at level L, is computed. The ratio 
image M has unity mean value. Due to imperfect despeckling, 
intrinsic signal variability of rough surfaces, and instability of 
image ratioing, the spurious fluctuations in homogeneous areas 
around the mean value m = 1 appear obviously. 
By denoting with <j t the standard deviation of the distribution 
of the modulating SAR texture M, a soft thresholding with 
threshold 0 = k <j t is applied to M : Values outside the 
interval [1 -k cr t ; 1 +k <7 t ] are diminished in modulus by 
k (7 t t and retained, while values lying within this interval are 
set to the mean value. The resulting texture map M s contains 
spatial features which are easily detected by the SAR image. It 
can be properly modulated into wavelet planes of the 
panchromatic image produced by the a trous wavelet algorithm 
to obtain new wavelet planes. 
2.3 HPFM (high pass filtering based on modulation) fusion 
technique 
High pass filtering based on modulation fusion technique aims 
at preserving the spectral characteristics of the MS image by 
keeping the result independent of the spectral property of the 
higher resolution image used for modulation (Pohl and 
Genderen, 1998). The HPFM processing formula for the case of 
the multi-spectral images sharpening by means of a co 
registered high resolution Panchromatic image P is the 
following 
j)=MS k (i, j)+ K(i, j) . P„(i, j) (3) 
where MS k is the resampled k-th band of the MS image, P H (i, 
j) is high frequence details of high resolution Panchromatic 
image. In the proposed method, it is replaced by the wavelet 
planes of the panchromatic image, produced by a trous wavelet 
algorithm, modulated with the SAR texture. K(i, j) is decided 
by the following forma 
K(i J)= MS k (i, j)/P L (i, j) (4) 
where P L (i, j) is low frequence information of high resolution 
Panchromatic image, and P(ij)= P k 0 > j)+ Ph (*> j)- 
The proposed fusion scheme applied to obtain the fused image 
is shown in Figure. 1. 
MS 
Figure 1 The proposed fusion schedule 
3. EXPERIMENTS AND ANALYSIS 
The input dataset is composed of an ENVISAT SAR image of 
urban area of Wuhan, Landsat-5 TM images (band 3, 4, 5) and a 
SPOT Pan image. All the data have been accurately co 
registered to the SPOT Pan image, as shown in Figures 2(a), 2(b) 
and 2(c). Texture will be blurred in the fusion image if the 
accuracy of co-registration among the SAR image, the TM 
images and the SPOT Pan image is not high enough. Figure 3 is 
the fused images by proposed fusion method. The fused images 
are evaluated in visual examination and quantitative analysis. 
2(a) ENVISAT SAR image