Premalatha Balan 
  
Baseline estimates for all these methods are listed in Table 2, and flattened interferograms for these baseline estimates 
are shown in Figure 3. The flattened interferograms were unwrapped to derive InSAR DEMs for all baseline estimates. 
There were six DEMs, including the one derived for baseline estimate using orbital information as a first trial. Since the 
DEM generated for baseline estimate using orbital information was already map-registered and differenced from the 
reference DEM, the other five DEMs were registered to the reference DEM as explained above, and differenced from 
the reference DEM. The results are discussed in the following section. 
4 RESULTS AND DISCUSSION 
Accurate baseline determination is one of the most important factors in determining the quality of INSAR derived DEM 
or any InSAR product. The method used to estimate the baseline is very important as further processing based on the 
estimated baseline significantly affects the quality of the final product. In this study, baselines were estimated using 
orbital information, image offset parameters, fringe rate method and combinations of these methods, producing six 
flattened interferograms using six different baseline estimates as mentioned above. All the six flattened interferograms 
using these baseline estimates follow to some extent the actual terrain topography as depicted on the map, though they 
differ from each other (Figure 3). Close examination and inter-comparison of the interferograms shows a number of 
discrepancies in the height values from the interferogram (by counting the fringes) for a few selected locations and 
actual height values shown in the map for the same locations. When DEMs were generated from these interferograms, 
the differences became very obvious. DEMs generated using the baseline value from the fringe rate method to estimate 
both perpendicular and parallel components of the baseline (method 3 and method 4), contained discontinuities (holes) 
located near the centre of the image, irrespective of the fringe estimation window location. This may be due to an 
inaccurate parallel component estimated by fringe rate method, which is a known disadvantage of this method (reported 
in Gamma software documentation). These two DEMs were ignored for further comparison. The other four DEMs 
registered to the reference DEM and a common area was cut from all these DEMs. Four InNSAR DEMS were subtracted 
from the reference DEM to generate respective difference images. The difference images showed tilt in different 
directions. The degree of tilt can be arrived at by taking profiles along the locations of lowest value and highest values 
in the difference images. The profiles and the profile locations for each image are shown in figures 4 to 7. Since the 
profiles for tilt were taken along different directions and for different locations, cross sections of all four DEMs and for 
the reference DEM, for same location in all images (along East-West direction of the image) were taken and are shown 
in Figure 8. 
  
Figure 4(a) shows 
that the tilt in the 
DEM generated 
using orbital 
information was in 
an east-west 
Pixel Value 
direction, with lower 
value in the west and 
higher value in the 
east. The difference T Cistance meter 
values for this DEM 
(a) (b) 
were above zero as 
seen from Figure 
4(b), meaning that 
the datum of the InSAR DEM was below the datum of the reference DEM (mean sea level), with a tilt in east-west 
direction. This is evident from the cross section of the InSAR DEM in comparison with the reference DEM shown in 
Figures 8(a) and 8(e) respectively. For the cross section shown in the Figure 8(a), the height values in the INSAR DEM 
for the baseline estimate based on orbital information were 300m below the reference DEM value in the west-most 
location and 500m below the reference DEM in the east-most location. Note that the shape of the INSAR DEM agrees 
with the shape of the reference DEM if the noise and the hole in the INSAR DEM are overlooked. 
  
8 2000 4000 6089 8860 
  
Figure 4 Profile along the tilt of the DEM generated using baseline estimate based on 
orbital information. (a) profile location and (b) the profile in the tilt direction 
  
  
  
The InSAR DEM generated using a baseline estimate derived from image offset parameters showed a northwest- 
southeast tilt with the lowest value in the southeast location and the highest value in the approximately northwest 
location, as shown in Figure 5(a). The degree of tilt is almost same (Figure 5(b)) as that of DEM generated using 
baseline estimate based on orbital information (Figure 4(b)) but the direction of tilt is different. When we compare the 
cross section of the DEM, then the DEM with image offset parameters used for baseline estimation shows a lower level 
datum. The height value of the InNSAR DEM in the west-most location is 400m below the height value of the reference 
DEM and height value in the east-most location is 300m below the reference DEM (Figure 8(b) and 8(e)). In this case 
also, the shape of the DEM agrees with shape of the reference DEM, while there are noise and holes as in the previous 
case. 
  
34 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B1. Amsterdam 2000. 
  
pe + BA P P4 PT