this process. This procedure continues recursively until all 
e pu 3.2.2 Cartographic Raster Data - Overviews There sub-regions are homogeneous. 
are three types of cartographic raster data: scaled physical 
orners ofa arameters, thematic maps, and background maps e.g. In our quadtree structure the east and the west hemispheres are 
print, The D epoctively: gravity maps, geologic maps, and global treated as separate quadtrees. The first quad level covers a 
Tg image mosaics. We store scaled physical parameter raster whole hemisphere. This then divides into four 90? longitude 
sud Shor data as binary images, without compression to preserve and latitude quadrants, and so on. Figure 5 shows the 
O describe numeric accuracy. For background maps, which are to be used coverage of quadtrees down to different levels of resolution. 
for display purposes only, we store these as tiles and The dark lines are regions where the colour information is 
le level 2 compress them by approximately 20 times using JPEG still inhomogeneous at these respective levels. This actually 
rted (fig | (Joint Photographic Experts Group) compression. In this high-lights the boundaries between different geological 
to describ way it is possible to store the best currently available global units quite well. 
hii image mosaic of the whole of the surface of Mars (231m per 
Yel ina pixel or 1/256th of a degree, 2.5Gb in Sinusoidal projection; 
Batson, 1986) in only 130Mb. The storage of thematic map 
descripto data has special requirements, namely it must be compact, it 
e err must be easy to portray at any scale (generalization), and it 
the image must be possible to perform rapid geographical searches. 
; used ag 3.2.2.1 Storage of Geological Maps using 
ye ve Quadtrees: For our test data we used a digital geological 
Down: In map of Mars (Tanaka, 1988). We processed this as a 
hs mas longitude-latitude cylindrical projection. It is thus 
i available straightforward to locate points on this for a given longitude 
contains a and latitude. However it has the disadvantage that the scale 
(m/pixel) in the longitude direction becomes distorted as a 
LE, function of cos(¢), and it is not easy to portray at low 
9$. 
resolutions without either: performing averaging processes 
over several pixels on the original map, or storing multiple 
resolution versions of the map with the consequential 
increase in storage space. 
Therefore to solve these problems we convert digital 
geological maps into quadtrees (Cromley, 1992). These 
allow both multiple resolution and compact data storage. The 
quadtree algorithm that we have used is based upon the 
following approach: if a coloured image region with a size of 
21 * 21 contains different coloured pixels (geological units), 
then it is declared inhomogeneous, and the algorithm splits 
it into four square sub-regions each 21-1 * 20-1 in size. Fig 4 
shows a geological map of Mars which has started to undergo 
   
different quadtree resolutions: top left 16x16 pixels, top 
right 8x8 pixels, lower left 4x4 pixels, lower right 2x2 
pixels. 
For the display of quadtree infomation as a map, at a given 
resolution, this boundary information needs to be filled in 
using a generalization method. This is achieved in the 
following order. If, in the next level down, the tree contains 
a majority of one colour (geological unit) over another, then 
this is used to fill in the gap. If this is not possible, then the 
geological unit present with the highest rank (rank pre- 
assigned to geological units according to the percentage area 
coverage of each unit on the surface of the planet) is used. If 
this fails then dithering of geological units is applied to fill 
  
  
  
  
  
  
  
  
in the gap. 
top right: 
g Scanner Raw image (2048 x | Quadtree storage for Quadtree storage 
2048) - no header colour + x,y position | for colour only 
4.19MB 3.08MB 0.51MB 
| necessary Tabel 5: Comparison of storage for a 2048 x 2048 pixel 
ect to the geological map of part of the western hemisphere of Mars. 
3 top left) 
nera (fig 3 3.2.3 Bibliographical References: During the Mars 
:amera (fig 96 mission, it is planned for co-investigators to have a 
In figure 2 "scientists note-book" facility whereby they can submit an 
pe used to email to the SYBASE SQL database system administrator, 
tions and concerning an image taken by the spacecraft. Keywords 
significant describing the Mars 96 image, are then added to the database. 
sition and This facility could be extended to include books, journals, 
jn is used. : : P : and conference proceedings, whereby certain images, named 
se moving im e ERE of 1e Pian of the, geological map cartographic OC and otre his on can be 
handle in d e sirüciure. included. Such a bibliographic list held within the database 
is unlikely to become fully comprehensive, however it will 
191 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996