All the images must be taken under similar illumina- 
tion conditions so as to avoid shadow differences ap- 
pearing along mosaic seams. The stereo illumination 
difference measure should not exceed 10 units. 
If possible, select images taken from the same orbit 
or closely spaced in time so as to avoid localised 
atmospheric effects associated with dust or clouds. 
If possible, choose images that have already been 
radiometrically corrected. 
All images should have similar view angle conditions, 
preferably near nadir so as to avoid parallax effects 
showing up near seams. 
The ratio of largest and smallest IFOV resolutions of 
the images used must not differ by more than two. 
Select images which are not previously known to be 
of poor quality. 
The selection criteria for images used for the production of 
colour composites are similar to those used in the mosaic- 
ing. One difference though is that we wish to find three 
images which overlap but which were taken through differ- 
ent filters. The majority of Viking Orbiter images acquired 
by the EXODUS project appear to have been taken in red 
and violet light, along with a smaller number of blue, mi- 
nus blue and green filter images. As a result of this, any 
colour composite produced will almost certainly be a false 
colour where the blue colour contribution is represented 
by a violet image, the green colour contribution is repre- 
sented by blue, minus blue or green, and the red colour 
contribution by itself. Another selection criterion was that 
allthree images should overlap each other by at least 10% 
of the total image coverage area. 
3.3 Selection of images for shape from shading 
MIS has been used to select images for shape from shad- 
ing work (Day et al, 1992). The requirements for selecting 
an image for this type of work are that their camera po- 
sition and orientation should be accurately known, and 
the images must be radiometrically corrected. MIS has 
also been used to select a suitably close range image of 
Phobos on which shape from shading was successfully 
applied (Murray et al, 1992). 
3.4 Selection of images for temporal studies 
Most features on Mars were imaged by the Viking Orbiters 
from many different illumination angles. MIS can be used 
to produce a list of images of a given feature which is 
sorted by solar azimuth. Such sequences of images, if 
taken from near nadir positions can be used for BRDF 
studies (Muller et al, 1992) and also for estimating relative 
heights from shadow length measurements. Also, images 
taken under similar illumination conditions, but sorted by 
time can be used for studying changes in the polar caps 
and the development of dust storms. 
3.5 Geographical query 
The example below illustrates the use of Prolog primitives 
for asking the following geographical query: find "mod- 
erately sized" craters on Mars which lie in "knobby hum- 
mocky terrain", and which are below 1km in altitude. The 
792 
“fail” on the end of the query is used to force it to back 
track and write out alternative solutions. e.g. 
| ?- crater(Name), 
moderate.size feature(Name), 
geology-at-feature(Name,'knobby hummocky terrain’), 
height.at feature(Name, Height), Height « -1.0, 
nl,write(Name), 
fail. 
Kirsanov 
Nutak 
Zuni 
If we now wished to see what images were available for 
the crater Kirsanov, these can be found as follows: 
| ?- list. images. by .feature('Kirsanov"). 
Image: 227S33 Resolution: 312.6 nvpixel 
Image: 239S54 Hesolution: 356.8 nvpixel 
Image: 247S32 Resolution: 364.7 mvpixel ... etc 
4. CONCLUSION 
The Mars Information System described in this paper has 
been successfully used during the EXODUS project to se- 
lect images suitable for automated stereo matching. In 
addition, it has been demonstrated to be a very flexible 
system for investigating what image coverage of Mars is 
available at UCL for several other applications which in- 
clude mosaicing, shape from shading, colour composites 
and examples of specific Martian processes. The high 
level nature of MIS has encouraged its use, and as a 
result new queries are frequently being put to it by the 
members of the EXODUS team. 
It is planned that MIS will control an end to end system for 
DEM production from Viking Orbiter and Mars Observer 
images using stereo matching and shape from shading 
techniques (Muller et al, 1992). In particular it will coor- 
dinate the selection of images and control points used to 
form a photogrammetric block of a given region of Mars, 
and also to record the storage and processed state of im- 
age and DEM data. 
Acknowledgements 
The EXODUS project is supported by SERC grants GR/F- 
84-294 and GR/F-82-894. We would also like to thank 
Chris Jones and the Information Technology Center at the 
Polytechnic of Wales, for use of their computing resources 
during the initial development of an experimental Prolog- 
based system which gave the insight into MIS. The follow- 
ing members of the USGS, Flagstaff have kindly supplied 
us with image position and orientation geometry, control 
point, and DEM data: Sherman Wu, Debbie Cook, Kay 
Edwards and Annie Howington. Wei Xia of the Goddard 
Space Flight Center has given some helpful advice on the 
photogrammetry of Phobos. 
References 
Blasius, K.R., Vetrone, A.V., and Martin, M.D., 1980, 
Viking Orbiter Stereo Imaging Catalog, NASA Con- 
tractor Report 3277, NASA.