ISPRS, Vol.34, Part 2W2, “Dynamic and Multi-Dimensional GIS”, Bangkok, May 23-25, 2001 
The original image of test site is shown as Fig.3. The result is 
shown as Fig 4. 
3.2 multi-route integral processing 
There is obviously degrade from speckle existing in the 
original image. The noise filtering can not applied in the 
elevation image shown as Fig. 4. The result of integral of Sh 
along close route should be zero, as follows: 
y = <j‘f(x)dx (8) 
L 
The multi-route integral of Sh can be used to smooth the 
noise appearing in the elevation map as Fig. 4. The result of 
multi-route process is shown as Fig 5. 
3.3 Improvement with the relief context 
It may be easily verified that speckle filtering doesn’t 
contribute efficiently to noise removal. We try to decrease both 
effect from approximation by /3=0 and speckle noise through a 
contextual relaxation using simulated annealing. Two formulas of 
each pixel’s energy are deduced as Equation (9) and (10): 
U(x) = U,(x)+U y {x) <*> 
(10) 
4. Conclusion 
The result, which has been obtained with this 
reconstruction technique of the relief, is promising. With regard 
to the contour map extracted from a digital elevation map, the 
morphological quality of the relief has been correctly restituted. 
However, the principal problems, like the vegetal cover of the 
ground and local character of the hypotheses have to be 
improved further. 
5. Acknowledgement 
The work discussed in this paper was supported by the 
HKSAR Government RGC grant (CUHK 4434/98E) and the 
Natural Science Foundation of China (No. 69782001). 
Reference 
Gao, Wen, Chen Xilin, 1999, Computer Vision-algorithm and 
System Theory, Tsinghua University Press, Beijing 
Horn, B., 1975, Obtaining Shape from Shading Information, 
McGraw-Hill 
Kang, Lishan, Xie Yun, 1994, Non-numerical Parallel Algorithm, 
Science Press, Beijing 
LEBERL, Franz W, 1990, Radargrammetric Image Processing, 
Artech House 
Taud, Hind and Jean-Francois Parrot, 1995, Qualitative digital 
elevation model from SAR data, SPIE Proceedingsof Synthetic 
Aperture Radar and Passive Microwave Sensing, Vol. 2584, pp. 
226-233 
di) 
Toutin, Thierry, 2000, Laurence Gray, State-of-art of elevation 
extraction from satellite SAR data, ISPRS Journal of 
Photogrammetry & Remote Sensing, Vol. 55, pp. 13-33 
Then, we apply the markovian strategy on the radar image Fig 5, 
which is the elevation map of multi-process. The obtained result 
is shown as Fig. 6. It shows that the strategy for smoothing the 
noise is promising. 
Wildey, Robert L., 1986a, Radarclinometry, Earth, Moon and 
Planets, Vol.36, pp.217-247 
Wildey, Robert L., 1986b, Radarclinometry for Venus Radar 
Mapper, Photogrammetric Engineering and Remote Sensing, 
Vol.51, No.1, pp.41-50