International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004 
0 ” 
a. The ideal phase og 
  
  
b. The practical phase 
Figure! The ideal phases and the practical phases 
Most of the phase unwrapping methods are excellent 
because of their comprehensive and complex considerations. 
But the existence of noises usually makes them be unsuccessful 
for all INSAR data. If there is no noise, the phases can be 
unwrapped successfully even by the easiest and simplest 
Algorithm. So, this paper pays more attention to the filtering of 
interferogram before the processing of phase unwrapping, and 
the successful phase unwrapping for the interferogram filtered 
by vector filtering is realized by using region grow method 
based on coherence image and on fake coherence image. 
  
  
Interferogram 
| 
Vector Filtering 
Coherence Image 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
————————-- 
Fake | 
Coherence | 3 k 
Image Filtered Interferogram 
= 
| 
  
Y 
  
The Region 
defined by 
Region Grow 
Y 
  
  
  
Phase Unwrapping 
d 
Unwrapped Phases 
  
  
  
  
  
  
  
  
  
  
Figure2 The Flow of this phase unwrapping scheme 
3. THE PHASE UNWRAPPING ALGORITHM USED IN 
THIS PAPER 
This paper presents the argument that the interferogram 
filtering is a key technology in phase unwrapping because the 
filtering can make it easy to succeed. So, a new phase 
unwrapping schedule is presented with the idea. The 
processing steps are shown in Figure 2. It includes two steps: 
the vector filtering and the region grow phase unwrapping. 
102 
Firstly, the interferogram is filtered to high quality by the 
method of vector filtering. Secondly, the filtered 
interferogram is unwrapped successfully by selecting a 
compatible gate value of coherence or of fake coherence with 
region grow method. 
3.1 Interferogram Filtering 
The filtering of interferogram has great influence on phase 
unwrapping, so this paper adopts the vector filtering method, 
which is better than others. Several common filtering methods 
for interferogram were analysed and compared in order to 
show the advantages of vector filtering. 
There are three interferogram filtering methods: 
non-adaptive filtering method, adaptive filtering method and 
multilooking filtering method. 
| Non-adaptive filtering methods These methods 
include mean filtering, median filtering, and so on. All of 
them are effective for optical images, but not for INSAR 
interferograms. These methods adopt some fixed windows for 
filtering and can induce the distortion of phases due to the 
periodicity character of interferogram not being considered. 
In order to show the cases, some phases (they are presented 
by 256 gray levels) in the window sized 3 by 3 at the fringe 
edge are shown in Figure3a. If we adopt the mean filtering, 
the phase of the pixel in center window is calculated: 
8-5È La, -78- Where, gi; is the phase of pixel 
i=l zl 
(ij ). 
But the phase considered the periodicity of the fringe 
should be calculated by the phases modified according to the 
ati id (The 
oif 
feature of interferogram, that is: Le = 
iE 
i=] j=l 
phases modified are shown in Figure3b). 
  
  
  
  
  
  
  
  
  
  
  
114 16 
14 | 40 | 239 jd 
3. 1-0 .1 225 3.19 1-30 
77 | 26 | 13 742 | 18 
7 6 
  
  
  
  
Figure3a the raw phases Figure3b the phases modified 
In the same way, if we adopt the median filtering method, 
the result phase filtered is: g — median{g, } z 26 but 
if considered the periodicity, it should be: 
g = median{g’} =13 . In the above formula, 
median( } presents the calculation of getting the median 
phase in the filtering window. 
From the results of mean filtering and median filtering, we 
can see that they are not fit for the filtering of interferogram. 
LJ Adaptive Filtering method ( Jong-Sen Lee,1998) This 
method is aiming at the orientation of the fringe, and uses 
directionally dependent windows to filter noises along the 
fringes. It selects one of the 16 window plates, whose 
direction is most fit for the fringe. It's better than the above 
Non-adaptive filtering methods, but it requires phase 
unwrapping in the operating window before filtering and 
phase rewrapping after filtering. What's more, it has to do a 
lot of work to select the directional window plate. So it is not 
so effective for filtering of interferogram because of difficulty 
of phase unwrapping and its slow processing 
| Multilooking Filtering method It is a common 
Internatic 
filtering 
S = 
Where, 
the ma 
value « 
present: 
main ir 
pixels; 
phase. 
If we 
S = 
This 
and is u 
of the 1 
result c 
multi-lc 
and $, 
respecti 
the resu 
Ve 
interfer 
intensit 
amelior 
The 
T 
e 
Where 
(k.[)in