Premalatha Balan 
  
To compare the InSAR DEM and the reference DEM it is necessary for the InSAR DEM to be geo-referenced. 
Registering DEMSs is a difficult task as there are no well-defined ground control points in the smoothly undulating 
height values of the DEMs. In order to register the INSAR-derived to the reference DEMS, the amplitude image (which 
is in the same geometry as the INSAR DEM) was registered to the map to produce a registration polynomial between 
SAR geometry and map geometry. This registration polynomial is then used to transform the INSAR derived DEM to 
map geometry. As the amplitude image for the study area did not have many well-identifiable ground control points, the 
PAN image from the IRS-1D satellite was first registered to the map, since the PAN image had many well identifiable 
cultural features which are also shown in the map. The amplitude image was then registered to the PAN image. The 
PAN image and the SAR amplitude image had some well identifiable feature in common. In order to achieve the best 
possible interpretability, the coherence and amplitude images for both the SLC images were added to generate a colour 
composite image. This composite is registered to the PAN image that is in the map geometry. The registration 
polynomial developed to register the INSAR colour composite image to the PAN image is used to transform the InSAR 
DEM to map geometry. 
A common area between the registered DEMs was cut from larger images in order to compare height differences 
between the DEMSs. The registered InNSAR DEM is subtracted from the reference DEM to generate a difference image. 
This difference image showed systematic height differences, i.e., there was an increase in height difference from West 
to East of the image. This indicated a tilt in the INSAR DEM due to an inaccuracy in a processing parameter. 
3.2 Baseline estimation using other methods 
The systematic height difference could be removed by two ways. One is by estimating the tilt from the difference image 
and removing it by adding a tilt component to every pixel in the DEM. Alternatively, the reason for the tilt could be 
analysed. Determining the source of the tilt is more important than just removing it by estimating from the tilted DEM. 
Careful analysis of each step of the InSAR processing chain revealed that the tilt was due to inaccurate baseline 
estimation. 
The state vectors provided in the leader file do not appear to represent the orbital positions of the Shuttle very precisely, 
hence it is necessary to estimate the baseline from the image and/or interferogram parameters. Other methods of 
baseline estimation attempted are listed below. 
1. Orbital information was used to estimate both perpendicular and parallel components. 
2. Image offset parameters were used to estimate both perpendicular and parallel components. 
3. The fringe method was used to estimate both perpendicular and parallel components with default estimation 
window location (centre of the image) and default window size (512 in range and 1024 in azimuth). 
4. The fringe rate method was used to estimate both perpendicular and parallel components by choosing a 
window location over a relatively flat area with a window size that covers just the flat area chosen. 
5. The fringe rate method was used to estimate the perpendicular baseline component with estimation window 
used in method 4, and orbital information was used to estimate the parallel baseline component. 
6. The fringe rate method was used to estimate the perpendicular baseline component with estimation window 
used in method 4, and image offset parameters were used to estimate the parallel baseline component. 
  
  
  
  
Method Perpendicular Parallel 
component (m) | component (m) 
Orbital information is used for both the components 139.2467 -31.2750 
2. Offset information is used for both the components 143.7984 -40.7094 
3. Fringe rate method is used to estimate both the components when the 123.7560 -0.0060 
estimation window is at the centre with the default window size (512 in 
range and 1024 in azimuth) 
  
4. Fringe rate method is used to estimate both the components by placing the 150.2302 0.0500 
window over a flat area with a user defined window size. The area and 
the window size were selected by analysing the contour pattern in the 
map. 
  
5. Orbital information is used to estimate the parallel component and fringe 149.9380 -31.3780 
rate method is used to estimate the perpendicular component. The fringe 
estimation window was the same as the one used in method 4. 
  
6. Offset information is used for parallel component and fringe rate method 149.9669 -40.7688 
is used to estimate the perpendicular component. The fringe estimation 
window was the same as the one used in method 4. 
  
  
  
  
  
  
32 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B1. Amsterdam 2000.