The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008 
1059 
Once noise reduction was completed, the 3D point cloud of area 
of interest was extracted from the four HRSC intersection point 
datasets. As noise reduction was carried out based on the same 
reference terrain model (MOLA) and most of the errors 
occurred in the strip edge were removed, DTM mosaicking was 
accomplished by combining the extracted four point clouds 
directly. Profile comparison was then applied to gridded DTMs 
from three different sources, including a mosaiced DTM after 
noise reduction (UCL DTM), MOLA DTM and the latest DTM 
released by DLR (DLR V50 DTM) which is available from the 
ESA PSA. For DLR V50 DTMs, they were mosaicked with no 
feathering function (Figure 6). As DLR V50 DTM was in a 
gridded format, the other two DTMs were also gridded using 
the same ground resolution (75 m). Three profiles for 
comparison across hi 000 and hi Oil (line #1), hi Oil and 1022 
(line #2), and hi022 and h0103 (line #3) are indicated in Figure 
6. The comparison result is shown in Figure 7. 
Figure 5. Test area over Ares Vallis. 
Figure 6. DLR V0 mosaicked hill-shaded DTM (top). The 
following three rows show the profiles in DLR V50 DTM, 
MOLA DTM (orange points represent MOLA points) and UCL 
DTM (magenta points indicate updated intersection points). 
Profiles across h1000 and h1011 
Profiles across h1022 and h0103 
Longitude (degree) 
Figure 7. Profile comparison. Top to bottom: line #1, line #2 
and line #3. 
From Figure 7 it is seen that the discontinuity between adjacent 
DTMs (marked by the 2 magenta lines) was clearly shown in 
the DLR version 50 DTM, in which the height difference could 
reach up to 200 metres. While in the UCL DTM, the gap was 
improved due to the application of noise reduction. In addition, 
as the MOLA 3D heights were treated as ground truth and used