The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bl. Beijing 2008 
In this paper, we have analyzed the principle and distribution 
model of speckle and proposed a speckle denoising algorithm 
based on bsf and DT-CWT. The bsf is used to induce the model 
of the dependencies of wavelet coefficients of the adjacent 
scales, and the noise joint pdf is reduced. Local variance 
estimation and wiener filter are used to estimate the filtered 
image. Experiment results with real SAR images have shown 
that much more speckle is removed and edges feature of 
de-noised images keep smooth and continuous. But four bsf 
models in (Levent S.endur et al., 2002a) are designed for optical 
images. The definition of the bsf model for SAR images is the 
future work. 
REFERENCES 
David Donoho L, 1995. De-noising by soft-thresholding, IEEE 
Transactions on Information Theory, 41, pp. 613-627. 
Levent S.endur and Ivan W. Selesnick, 2002a. Bivariate 
shrinkage functions for wavelet-based denoising exploiting 
interscale dependency, IEEE Transactions on Signal Processing, 
50(11), pp. 2744-2756 
Levent S.endur and Ivan W. Selesnick, 2002b. A bivariate 
shrinkage function for wavelet-based denoising, Proceedings of 
IEEE International Conference of Acoustics, Speech, and Signal 
Processing. 2, pp. 1261-1264. 
Levent S.endur and Ivan W. Selesnick, 2002c. Bivariate 
shrinkage with local variance estimation, IEEE Signal 
Processing Letters, 9(12), pp. 438-441. 
Marc Simard, Gianfranco DeGrandi, Keith P. B. Thomson, and 
Goze B. B'eni'e, 1998. Analysis of speckle noise contribution 
on wavelet decomposition of SAR images, IEEE Transactions 
on Geoscience and Remote Sensing, 36(6), pp. 1953-1962. 
Nick Kingsbury and Julian Magarey, 1997. Wavelet transforms 
in image processing, Proc. First European Conference on Signal 
Analysis and Prediction, Prague, pp. 23-34 
Nick Kingsbury, 1998a. The dual-tree complex wavelet 
transform: a new technique for shift-invariance and directional 
filters, in DSP Workshop 
Nick Kingsbury, 1998b. The dual-tree complex wavelet 
transform: a new efficient tool for image restoration and 
enhancement, Proc. European Signal Processing Conference, 
EUSIPCO 98, Rhodes, pp. 319-322. 
Peter de Rivaz and Nick Kingsbury, 2001. Bayesian image 
de-convolution and de-noising using complex wavelets, Proc. 
IEEE Conf. on Image Processing, Greece, paper 2639 
Wang Hongxia, Cheng Lizhi and Wu Yi, 2005. A complex 
wavelet based spatially adaptive method for noised image 
enhancement, Journal of Computer Aided Design & Computer 
Graphics, 17(9), pp. 1911-1916. 
Xiao Guochao, Zhu Caiying, 2001. Radar Photogrammetry, 
Earthquake Press, Beijing. Pp.56-58 
Yang Mengzhao, Li Chaofeng and Xu Lei, 2005. Image 
denosing based on complex wavelet transform and H-Curve 
criterion, Computer Applications, 25(4), pp. 769-774. 
Yi Xiang and Wang Wei-ran, 2004. The application of statistic 
model for complex wavelet coefficients in image denoising, 
Opto-Electronic Engineering, 31(8), pp. 69-72. 
Yi Xiang and Wang Weiran, 2005. A probability model for 
adaptive image denoising, Acta Electronica Sinica, 33(1), pp. 
63-66. 
Zhang Chunhua, Xu Suqin, Tao Ronghua and Chen Biao, 2005. 
Analysis of speckle reduction of sar images based on 2D 
complex wavelet, Journal of Institute of Surveying and Mapping, 
22(4), pp.269-271. 
Original image 
Image denoised by 
bsf with DWT in 
(Levent S,endur et 
al., 2002a) 
Image denoised by 
bsf with DT-CWT 
in (Levent S,endur 
et al., 2002a) 
Image denoised by 
bsf with DWT in 
this paper 
Image denoised by 
bsf with DT-CWT 
in this paper 
PNSR 
ENL 
PNSR 
ENL 
PNSR 
ENL 
PNSR 
ENL 
PNSR 
ENL 
Figure 3 (a) 
12.8174 
7.3667 
12.8559 
7.492 
13.0292 
8.3072 
13.3294 
12.5415 
13.432 
20.501 
Figure 3 (b) 
11.3375 
3.8817 
11.3812 
3.9887 
11.5758 
4.5674 
12.2223 
8.8573 
12.3557 
14.8015 
Figure 3 (c) 
10.2375 
0.40142 
10.9974 
0.48088 
11.0751 
0.48205 
11.7579 
0.79368 
12.2304 
1.3207 
Figure 3 (d) 
7.8752 
0.17839 
8.0492 
0.18679 
8.2887 
0.18828 
9.218 
0.25724 
9.7728 
0.29422 
Figure 3 (e) 
3.4502 
45.594 
3.6175 
101.9069 
3.6309 
264.2059 
3.7994 
218.1747 
3.7732 
509.6754 
Figure 3 (f) 
3.7318 
1.8667 
3.9239 
3.2425 
3.9489 
7.5523 
4.0607 
6.3549 
4.05 
22.146 
Figure 3 (g) 
31.3577 
34.938 
32.8672 
34.2168 
33.6336 
34.0187 
33.5806 
49.7928 
33.8201 
57.7728 
Figure 3 (h) 
23.6098 
80.8072 
25.2745 
138.3005 
26.801 
151.9933 
26.6296 
257.5009 
27.6259 
300.5978 
Table 1. 
Statistics of PNSR and ENL of figure 3 
161