The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bl. Beijing 2008 
121 
pseudo-edges have disappeared (see region marked by A), 
while the true edge features have been efficiently preserved and 
the false edges are eliminated (see regions marked by B and C). 
W 
Fig.3 The edge detection result of original image (a) and the 
image filtered by our algorithm (b) 
1) The standard deviation, RSD of our filter is obviously 
smaller than the traditional filters. The standard deviation, RSD 
of our filter is obviously smaller than the traditional filters. This 
means the ability of denoising is better than that of them. 
2) The PSNR value maintains the same level with other filters, 
which represents the quality of filtered image is similar. But the 
ENL, EPI of proposed method is obviously better than the other 
filters. It means that our filter can not only efficiently reduce 
the speckle noise but well preserve the detail in the image. The 
main reason is that in our method before filtering we acquire 
the robust edge information through wavelet transform modulus 
maximum algorithm and edge fusion, which is then used as 
guidance for speckle reduction. 
3) Fig.3 is the edge detection results in original and filtered 
image. It can easily find that by speckle reduction by our 
method many false edges have been disappeared while the main 
structures of the objects in the image are retained. It can also 
express that our method can efficiently reduce the speckle and 
well preserve the edge in the image. 
6. CONCLUSION 
Existing speckle filters can effectively reduce speckle effects 
but unfortunately also lose image details. In this paper, we 
propose a wavelet transform speckle reduction algorithm for 
SAR imagery based on edge detection. Through wavelet 
transform modulus maximum algorithm and edge fusion, this 
guarantee the edge information obtained is robust. This is then 
used to preserve the edge while filtering. Experiments have 
been performed and the filtering results of our filter and other 
traditional filters have been elaborately analyzed. From the 
results, we go to the conclusion that our method can not only 
efficiently reduce the speckle noise but well preserve the edges 
in the image. 
ACKNOWLEDGEMENTS 
Thanks for the supporting from the 973 Program of the 
People’s Republic of China under Grant 2006CB701302 and 
the National Natural Science of China under Grant 407721001. 
Through analyzing the results in Table 2, we can go to the 
following conclusion: 
Images 
Mean 
Deviation 
RSD 
ENL 
PSNR (dB) 
EPI 
Original image 
70.9431 
23.2387 
0.3276 
9.3196 
— 
1 
Proposal method 
69.859 
18.366 
0.2629 
14.469 
28.230 
0.6149 
Lee 
70.826 
19.210 
0.2712 
13.638 
27.076 
0.5502 
Gamma 
70.709 
19.113 
0.2703 
13.778 
27.709 
0.5420 
Median 
69.793 
19.041 
0.2728 
13.882 
27.327 
0.5736 
Frost 
71.067 
19.573 
0.2754 
13.183 
27.955 
0.6082 
Table 2. Comparison results of our filter with other traditional filters 
REFERENCES 
Crimmins T. R. 1986, Geometric filter for reducing speckle, 
Optical Engineering, vol.25, pp. 651-654. 
Dong Y., Forster B. C., Milne A. K., Morgan G. A. 1998, 
Speckle suppression using recursive wavelet transforms, 
International Journal of Remote Sensing, 19(2), pp.317-330. 
Dong Y., Milne A.K., Forster B.C. 2001. Toward edge 
sharpening: a SAR speckle filtering algorithm. IEEE 
Transactions on Geoscienceand Remote sensing, 39(4), 
pp.851-863. 
Donoho D.L. 1995. Denoising by soft-threshoding. IEEE 
Transactions on Information Theory, 4(3), pp.613-627.