inbul 2004 
0. 3, table 
; the value 
ages with 
geometric 
he images, 
on, further 
on for the 
jared with 
e a known 
nage rays 
way, the 
mputation 
t the two- 
id 338m3. 
e 4). 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004 
  
Zus Z ta, Mi Zu =Z, ) (7) 
where Z,, mean value of the manually measured DTM 
Lyi DTM-height no. i of the manually measured DTM 
Zi DTM-height no. i of the destination DTMy 
  
  
| rms [m] | 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
140.1 
  
d images 
para meters 
334m2, 338m3 
n of image 
;tor 2x) 
, we have 
ese DTMs 
{ap and a 
a way that 
n the two 
in section 
sht change 
ent initial 
Lommel- 
tial DTMs 
v, that the 
our pixels 
metres. It 
rect result 
je average 
  
  
m scale factor 
do height offset 
. DTM parameter — | — Analysis and accuracy parameter 
‘Scale factor m | Offset ay [m] | Iterations Zeim] | nus[m] — 
Ta 0.0 4 -192 140.1 
0.5 0.0 9 -84 [463 
Ten 0.0 7 9.5 150.5 
F 1 360.0 6 36.3 148.4 
0.5 360.0 18 14.2 145.4 
1 720.0 12 35.4 148.4 
BS 720.0 ZF" 6 -| 1431 4 
TS 1080.0 18 16.5 142.3 
0.5 1080.0 16 214 143.6 
RE 1440.0 29 11.6 AA ea 
edit he: 1440.0 26 i1 LO d 11229. 
nl 1800.0 17 1801.9 1807.9 
0.5 1800.0 83 14.5 147.0 
] 2160.0 10 21845...|] 21902 
0.5 2160.0 18 2174.5 2181.4 
  
  
  
  
  
  
  
Table 5. Radius of convergence using two images 
4. CONCLUSIONS 
The represented work on MI-SFS shows that the derivation of a 
high-resolution DTM of real digital planetary images by means 
of MI-SFS is feasible. The one-image and multiple-image 
analyses are carried out using imagery from the lunar mission 
Clementine. The obtained results shows that MI-SFS is a 
method which is able to close the gaps in DTMs determined 
with other reconstruction methods. Furthermore we show, that 
the required initial values have a radius of convergence of about 
four pixels (in this case of about 1440 metres). 
The computation with one of the images was not successful. In 
this case we developed a simple, and successful method which 
modified the observed grey values. In order to do so, we also 
needed an initial-DTM of the area, and although we have not 
yet explicitly checked this assumption, we believe that we can 
compute the grey value modification based on the DTM which 
serves as initial height values for the whole approach. 
In future we will intensify our investigations to simultaneously 
process two and more images within MI-SFS. We will also try 
to increase the geometric accuracy by introducing more 
sophisticated object surface models (one times one pixel DTMs 
with appropriate smoothness constraints, breakline and 
occlusion detection modules). We also plan to integrate the line 
sensor geometry into the algorithm, to use other planetary data, 
€. HRSC data. In addition, the next important step is the 
combination of image matching with MI-SFS into a combined 
method. A precondition for such a combination is a separation 
of the surface under consideration into parts with constant 
albedo (MI-SFS) and into parts with variable albedo (image 
matching). This task remains a challenge of the whole approach 
Which we will try to tackle using texture analysis. 
WI 
5. REFERENCES 
Dorrer, E., 2002. Digital Elevation Model Refinement by 
Shape-from-Shading. Bildteknik/Image Science, Stockholm, Nr. 
2002:1, pp. 57-67. 
Fua, P. Leclre, Y. G., 1995. Object-Centered Surface 
Reconstruction: Combining Multi-Image Stereo and Shading. 
International Journal of Computer Vision, 16, pp.35-56. 
Gaskell, R. W., 2003. Optical only determination of small body 
shape and topography. ISPRS Extraterrestrial Mapping 
Workshop "Advances in Planetary Mapping 2003”, Houston. 
Hapke, B., 1993. Theory of Reflectance and Emittance 
Spectroscopy, Topics in remote sensing Vol. 3, New York, 
Cambridge University Press. 
Heipke, C., 1992. Integration of digital image matching and 
multi image shape from shading. /SPRS, Washington, Vol. 
XXIX, Part B3, pp. 832-841. 
Heipke, C.  Piechullek, .C., . 1994. Towards surface 
reconstruction using multi image shape from shading. /SPRS, 
Vol. XXX, Part 3/1, pp. 361-369. 
Horn, B. K. P., 1986. Robot Vision. The MIT Press, 509 p. 
Horn, B. K. P., Brooks, M. J., 1989. Shape from Shading, The 
MIT Press, 577 p. 
Lohse, V., Heipke, C., 2003. Derivation of digital terrain 
models by means of multi-image shape-from-shading: results 
with Clementine images. /SPRS Workshop “High Resolution 
Mapping from Space 2003”, CD-ROM, University of Hannover. 
McEwen, A. S., 1991. Photometric. Functions for 
Photoclinometry and Other Applications. Icarus 92, pp. 298- 
311: 
Nozette, P. et al., 1994. The Clementine mission of the moon. 
Science, Vol. 266, pp. 1835-1839. 
Piechullek, C., 2000. Oberflächenrekonstruktion mit Hilfe von 
einer Mehrbild-Shape-from-Shading-Methode. Deutsche 
Geodätische Kommission, Reihe C, Heft 518, 96 p. 
Rebhan,  H. 1993. Richtungsabhingige 4 Reflexions- 
eigenschaften der lunaren Oberfläche. Dissertation, München, 
DLR-Forschungsbericht 93-28. 
Smith, D. E., et al, 2001. Mars Orbiter Laser Altimeter: 
Experiment summary after the first year of global mapping of 
Mars. Journal of Geophys. Res., 106 (E10), pp. 23689-23722 
Zhang, R. et al., 1999. Shape from Shading: A Survey. /EEE 
Trans, PA&MI, Vol. 21, pp. 670-706. 
6. ACKNOWLEDGEMENT 
This work was developed within the priority program "Mars 
and the terrestrial planets" financed by the Deutsche 
Forschungsgemeinschaft (DFG) under the project number HE 
1822/10. The support is gratefully acknowledged. Thanks also 
go to DLR for providing the Clementine data, and to the 
employees of the ISIS Support Center at the United States 
Geological Survey (USGS) for their aid in the radiometric 
calibration of the Clementine images.