5.3 Parallax Measurements 
How well can the parallax difference be measured? 
Experiments with visual repeat measurements provide 
confidence that an operator can define the stereo matches to 
within +0.6 pixels. Automated matching may differ from 
visual observations with an r.m.s. error of +2 pixels (Leberl et 
al., submitted). 
5.4 Future Work 
Hopes to quantitively assess Magellan-stereo accuracies are 
based on the analysis of internal mismatches in overlapping 
stereo-models at higher latitudes, on assessing elevation 
differences from stereo and from exploiting symmetric features 
in single images (Leberl et al., 1991). 
5.5 Vertical Exaggeration 
The parallax-to-height conversion in photogrammetry typically 
is by a factor of about 0.6 (the so-called base-to-height ratio). 
Therefore a 100 m photographic parallax will be caused by a 
167 m terrain elevation difference. 
In Magellan's radar stereo images, this is a ratio of 1.4 or more: 
a 100 m parallax will be caused by only a 60 m elevation 
difference. Therefore we find that Magellan stereo images can 
have very large stereo parallaxes. An error of parallax of 1 
pixel will result in an error of elevation of only 0.6 pixels. As a 
result we have a surprisingly strong vertical expression in 
Magellan SAR stereo data, which is stronger than conventional 
mapping photography could produce! 
6. CONCLUSION, OUTLOOK 
Magellan has created the largest stereo radar data set ever. The 
vertical expression of relief in these stereo-models is stronger 
than it would be in mapping photography, and stronger than in 
past air or spacecraft stereo radar on Earth. 
We are developing tools to process these data with rigorous 
radargrammetric methods based on state and velocity vectors 
of the spacecraft and a sensor model of the SAR. In the interim 
demonstration products are being generated with simplified 
algorithms to review the capabilities of Magellan as a stereo 
mapping system. We believe that accuracies of parallax 
measurement are about +1 to +2 pixels, and accuracy of 
terrain elevation can be in the range of +100 m. 
ACKNOWLEDGMENT 
Many people participate in a complex space mission, and some 
of them had to make an extra effort to provide experimental 
data. We are particularly grateful to Graig Leff at JPL for 
tracking down specific data. George Arnold was helpful in 
getting stereo measurements done, and Matt Jackson kept the 
stereo viewing software running smoothly. 
RADARGRAMMETRIC PUBLICATIONS ABOUT 
MAGELLAN 
Leberl, F., K. Maurice, J. Thomas, M. Millot, (Submitted), 
Automated Radar Image Matching Experiment. ISPRS J. of 
Photogrammetry and Remote Sensing. 
Leberl, F., K. Maurice, J. Thomas, W. Kober (1991) 
Radargrammetric Measurements from the Initial Magellan 
Coverage of Planet Venus. Photogrammetric Engineering and 
Remote Sensing, Vol. 57, No. 12, pp. 1561-1570. 
Leberl, F., K. Maurice, J. Thomas, C. Leff, S. Wall (in print), 
Images and Topographic Relief at the North Pole of Venus. J. 
Geophysical Research. 
Leberl, F. W., J. K. Thomas, K. E. Maurice (in print), Initial 
Results From the Magellan Stereo-Experiment. J. Geophysical 
Research. 
Leberl, F. W., K. E. Maurice, J. K. Thomas (1992) 
Radargrammetric Analysis With Magellan Data of Planet 
Venus. Proc. 58th Annual Convention, American Society of 
Photogrammetry and Remote Sensing Conference, 
Albuquerque, NM, pp. 253-263. 
Thomas, J., W. Kober, F. Leberl (1991) Multiple Image SAR 
Shape-from-Shading. Photogrammetric Engineering and 
Remote Sensing, Vol. 57, No. 1, pp. 51-59. 
800