Istanbul 2004
quality by the
the filtered
/ selecting a
oherence with
ence on phase
ering method,
ering methods
d in order to
ng methods:
g method and
1ese methods
so on. All of
X for INSAR
d windows for
es due to the
considered.
are presented
) at the fringe
nean filtering,
is calculated:
1ase of pixel
of the fringe
cording to the
5 gq47 The
-16
-30
13
jases modified
ering method,
} = 26, but
should be:
ve formula,
g the median
1 filtering, we
terferogram.
1998) This
ige, and uses
ses along the
lates, whose
ian the above
quires phase
filtering and
it has to do a
e. So it is not
e of difficulty
a common
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004
filtering method for interferogram. It is defined as follows:
n n
$2 € ble tel re [se^
mn yy =
Where, Y, (kD presents the complex value of the pixel in
the main image; V, (k = c.l + h) presents the complex
value of the corresponding pixel in the slave one; S|
presents the module of S. (c, h) is the offsets between the.
main image and the slave one according to the calculated
pixels; The window is sized 7/7 by n; ¢ is the filtered
phase.
If we symbol A = y, (k, I) : |v, (k - cl h) ‚then
1 m n :
s-L F5 Ae = sf
mn yq
This filtering method is much effective for fringe filtering
and is used more and more. But, we can find that if the energy
of the noise is much higher than that of others, the filtering
result can be affected to lower precision. The sketch map of
multi-looking filtering is shown in Figure 4. Supposing 4,
and $, are the phases with noises, whose energies are
respectively 4 and 4,,but 4, is much higherthan 4,,so
the result phase 4. will be closeto $,.
Vector Filtering This is a new filtering method for
interferogram fringe. It can avoid the affection of coherent
intensity and has the virtue of fast processing speed, as is the
1h
Apt P
75 x
up À 2 A ~A
ET dE
~~ uS rT
eT zum
= 3 E
Te Thee Ee
Sa
fhe j$2
Figure4 Multi-looking Filtering
amelioration of Multi-looking filtering.
The Vector Filtering is defined as:
AN
=
A B"
s m
^
A N e J #1
/ , 7
/ S
j eM Fan =
| Tt \ d
EL | SERT 7
| TET Jia
\ 3d $2
\ y.
A 2
LÉ
RE —
Figure Vector filtering
| n n
MN i=l [=]
iPr i 3
€ (shown in Figure 5)
Where, ¢ is the result phase; $, , is the phase of pixel
( kl ) in the filter window.
The Vector Filtering method can be comprehended in
another way. Firstly, the interferometric phase $, , is
mapped to a vector elfe . which can be expressed by
[Pkt 2
fa = COS 0. a J sin Pr. :
That is the interferogram is mapped to a cosine image and a
sine image. Secondly, the cosine image and the sine image are
filtered by mean filtering method with the same window size.
Finally, the phase filtered by Vector Filtering is calculated
using the following formula:
COSQ,, and SING, , :e
z d=0 550
2
M a=0.5 <0
2
¢= sen a>0,b>0
a
2x + etant a>0,b<0
a
T+ Sen > a<0
a
Where, a is the value of the pixel in the cosine image
filtered by mean filtering; D is the value of the pixel in the sine
image filtered by mean filtering.
The Vector Filtering not only can avoid the affection of the
energy, but also has the virtue of fast and high efficiency. So,
it's a better filtering method for the filtering of interferogram
fringe.
3.1 Region Grow Phase Unwrapping
The second step is phase unwrapping using region grow
method by selecting a compatible gate value of coherence or
of fake coherence, as is described as follows: A region is
grown based on the coherence image or the fake coherence
image by given a seed pixel, whose coherence or fake
coherence value is high. Then the phases of different pixels in
this grown region are unwrapped by the direct-recursive
algorithm. The phases in other regions are not dealt and
signed that they are not unwrapped.
The coherence image is calculated by the following
Formula:
here iq se srence are re«snectivelv
Where, po B the coherence, ZZ ute respectively
*
the complex value of the master image and of the slave one. Zi
is the conjugate of Z , .
Likely, the fake coherence can be calculated by the
f
m n
(03 Y cos 9. , Y GC > sin 4, y
formula: »' — \ i=l j=1
ic