ize of an 
el image 
MxN. 
| manual 
) for the 
> defined 
e of the 
:uracy of 
pep ), 18 
(7) 
Where / denotes the number of non-zero pixels in Æ . Ideally, 
the pep value of a perfect segmentation should equal zero, so 
the smaller pep is, the better the segmentation. Thus, pep 
indicates the quality of the image segmentation. In this paper, 
we use the percentage of correct pixels ( pcp ) to represent the 
accuracy of segmentation: 
pp =1— pep (8) 
Table I compares the efficiency and accuracy of the above 
methods for the SAR image. 
TABLE I ACCURACY AND EFFICIENCY COMPARISON 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B7, 2012 
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia 
  
  
Method a Ay Az Efficiency- Accuracy — 
Time (s) pcp (%) 
ALGI 1500 0.05 1.5 49.658 64.821 
ALG2 1500 . 0.05 1.5 10.261 81.143 
ALG3 2000 0.05 1.5 10.796 79.135 
ALG4 1500 0.05 TS 5.184 96.705 
  
  
Notes: ALG4 is the proposed method. 
From the above qualitative analysis (Experiment 1 and 2) and 
quantitative analysis (Table I), the following conclusions can be 
obtained: 
1) By using OTSU algorithm to initialize the level set contour, 
segmentation accuracy is improved and iterations can be 
reduced significantly (please see the figurel (c)(d) and 
ALGI,ALG2 in Table I). 
2) Comparing with single-scale segmentation, the multi-scale 
technique can produce better accuracy and efficiency (please 
see the figurel (d)(f), and ALG2, ALGA in Table I). In any case, 
our method (ALG4) often leads to a superior result. 
4. CONCLUSION 
In this paper, a novel SAR water extraction method integration 
multi-scale level sets and OTSU algorithm is proposed. 
Although experiments have testified that our method performs 
better than previous level set methods, much work remains to be 
done. The multi-scale analysis framework is a new component 
in the level set method for segmentation and there is still much 
work to do in the multi-scale analysis. Further, the Gamma 
distribution is used to represent the energy functional, because 
of the characteristics of the SAR image; in future work we will 
consider a more suitable energy functional for SAR images 
produced by different sensors. Finally, a threshold segmentation 
method was used here to initialize the level set function, and 
was shown to be useful for a binary segmentation; however, for 
multiphase segmentation, this technique may require further 
development. 
5. ACKNOWLEDGEMENTS 
This work was supported by National Key Fundamental 
Research Plan of China (973) (No.2012CB719906) and 
National Natural Fund of China (NSFC) (No.41101414, 
No.40901211, and No.61001187). 
6. REFERENCES 
Ayed, I. B., Vazquez, C., Mitiche, A., and Bellhadj, Z., 2005, 
Multi-region level set partitioning of synthetic aperture radar 
images, IEEE Transactions on Pattern Analysis and Machine 
Intelligence, 27(5), pp. 793-800. 
Chan, T. F. and Vese, L. A., 2001, Active contours without edges, 
IEEE Transactions on Image Processing, 10(2), pp. 266—277. 
Cook, R. Connell I. M. and Oliver, C. J., 1994, MUM 
Segmentation for SAR Images, Proceedings of SPIE, 2316, pp. 
92-103. 
Fjortoft, R., Lopes, A. and Marthon, P., 1998, An Optimal 
Multiedge Detector for SAR Image Segmentation, IEEE 
Transactions on Geoscience and Remote Sensing, 36(3), pp. 
793-802. 
Gao, G., Zhao, L. J., Zhang, J., Zhou, D. F. and Huang, J. J., 2008, 
A segmentation algorithm for SAR images based on the 
anisotropic heat diffusion equation, Pattern Recognition, 41, 
pp. 3035-3043. 
German, O. and Refregier, P., 2001, Edge location in SAR images: 
Performance of the likelihood ratio filter and accuracy 
improvement with an active contour approach, IEEE 
Transactions on Image Processing, 10(1), pp. 72-78. 
Horritt, M. S., 1999, A statistical active contour model for SAR 
image segmentation, Image and Vision Computing, 17, pp. 
213-224. 
Law, Y. N., Lee, H. K. and Yip, A. M., 2008, A multiresolution 
stochastic level set method for mumford-shah image 
segmentation, /EEE Transactions on Image Processing, 
17(12), pp.2289-2300. 
Lee, J. S., 1989, Segmentation of SAR Image, IEEE Transaction on 
Geoscience and Remote Sensing, 27(6), pp. 981-990. 
Li, C., Xu, C., Gui, C. and Fox, M. D., 2005, Level Set Evolution 
Without Re-initialization: A New Variational Formulation, 
Proceedings of the 2005 IEEE Computer Society Conference 
on Computer Vision and Pattern Recognition, 1, pp.1-7. 
Martin, P., Refregier, P., Gpidail, F. and Guerault, F., 2004, 
Influence of the noise on level set active contour segmentation, 
IEEE Transactions on Pattern Analysis and Machine 
Intelligence, 26(6), pp. 799-803. 
Oliver, C. J., Quegan, S., 1998, Understanding Synthetic Aperture 
Radar Images. New York: Artech House. 
Osher, S. and Sethian, J., 1988, Fronts Propagating with Curvature 
Dependent Speed: Algorithms Based on the Hamilton-Jacobi 
Formulation, Journal of Computational Physics, 79(1), pp. 
12-49. 
Otsu, N., 1979, A threshold selection method from gray-level 
histogram, IEEE Trans. on Systems, Man and Cybernetics, 
9(1), pp. 62-66. 
Silverira, M. and Heleno, S., 2009, Separation Between Water and 
Land in SAR Images Using Region-Based Level Sets. IEEE 
Geoscience and Remote Sensing Letters, 6(3), pp. 471-475. 
Xu, X, Li, D. and Sun, H., 2003, Multiscale SAR Image 
Segmentation Using A Double Markov Random Field Model, 
Proceedings of Seventh International Symposium on Signal 
Processing and Its Applications, 1, pp. 349-352. 
Zhao, H.-K., Chan, T., Merriman, B. and Osher, S., 1996, A 
variational level set approach to multiphase motion, Journal of 
Computer Physics, 127, pp.179—195.