With simulations, the interferometric geometry 
being known, we can generate interferograms for 
different constant elevations. Figure 14 shows a 
profile of the interferometric geometry. The fringe 
lines are the intersections of the terrain with 
surfaces of constant distance differences that are 
multiple of the wavelength (hyperbolic profiles) for 
the satellites positions. Getting a control point or 
some information about the height of one point in 
the interferogram is equivalent to defining which 
fringe line will be the reference one. After 
unwrapping it is obvious to get directly the absolute 
height of each point since we just have to intersect 
the initial interferogram with the set of constant 
height interferograms. For a given point only one 
constant height interferogram presents the same 
phase value than the initial interferogram. This 
gives the height for this point. 
Sref Fringe lines 
Z = Cste 
  
Hyperbole 
fig. 14: Interferometric altimetry restitution 
This method is computationally intensive but is 
very easy to set up to get coarse absolute height 
information for these simulated interferograms. 
With real interferograms we need to have a precise 
orbitography for both satellite positions. Figure 15 
presents a coarse partial result for which we have 
computed constant elevation interferograms every 
100 meters. To get a dense DEM we affected 
several phase values for each level in order to get 
almost a full partition of the elevation range. 
  
fig. 15: Coarse altimetry restitution with 100 
meters slicing 
Compared with the initial SPOT DEM this coarse 
altimetry restitution is within 50 meters accuracy. 
3. EVALUATION 
To evaluate and compare DEM generated by 
interferometry we use SPOT stereogrammetry. 
Under a contract with C.N.E.S. we worked on a 
couple of SIR-B images taken over northern 
Canada. After phase-free preprocessing to obtain 
90 
complex slant range imagery from which 
deformations images were deduced, interferogram 
has been produced at C.N.E.S. (MASSONNET, 
1991). Figures 16 and 17 represent the module of 
the interferometric image in the reference 
geometry and the generated interferogram. 
  
fig. 16: Module image for reference SIR-B scene 
This particular interferometric data set was 
obtained with crossed orbits (GABRIEL, 1988) and 
preprocessing of the interferogram was performed 
at C.N.E.S. to reduce unwrapping ambiguities. The 
specificity of these data along with inaccurate 
informations about the shuttle orbits, encouraged 
us to simply address the unwrapping problem 
without trying to restitute the absolute altimetry of 
the scene. This could be done by adding to the 
unwrapped phases the phases laws removed by 
the preprocessing (MASSONNET, 1991) and by 
obtaining the absolute height from the orbit 
characteristics of the shuttle. Nevertheless without 
these informations we can still try to evaluate 
qualitatively the unwrapped phases: the 
transformation of these phases to be left to obtain 
DEM being locally linear transformations. 
With the region method described previously we 
partially unwrap the initial interferogram to get the 
result presented figure 18. 
  
fig. 17: Phase interferogram image 
Although this unwrapped result does not 
correspond to a digital elevation model it contains 
all the potential relative accuracy of this 
interferometric data set. We can then compare its 
relative altimetry values with a reference DEM and 
In particular a SPOT DEM (figure 19). Since the 
interferometric product is still in radar slant range. 
geometry, we have to set up the same geometry for