Académique Documents
Professionnel Documents
Culture Documents
Authorized licensed use limited to: UNIVERSITY TECNOLOGI MARA. Downloaded on December 29, 2009 at 15:21 from IEEE Xplore. Restrictions apply.
if the number of looks of the original SAR data is small Var( x)
(less than 4 looks), 2) the highly reflective point targets are with b , (2)
blurred and their powers reduced, and 3) dark spotty pixels
Var( z)
are not filtered.
Var ( z ) z 2K 2
Recently, Lee et al [1] improved the original Lee sigma and Var ( x ) v (3)
filter by eliminating these deficiencies. The bias problem is (1 K ) 2
solved by redefining the sigma range based on the speckle v
probability density functions. To mitigate the problems of
blurring and depressing strong reflective scatterers, a target In (1), (2), and (3), the local mean z and the local
signature preservation technique is developed. In addition, variance Var (z ) are the mean and the variance of pixels in
we incorporate the minimum mean square error estimator
the sigma range, respectively, and Var ( x ) is the variance
for adaptive speckle reduction.
of x , which is computed by (3). The parameter Kv is the
3.1. Bias compensation and dark pixels removal standard deviation of multiplicative noise of v , and is a
function of number of looks. For example, for 1-look
For the single look amplitude and intensity SAR data, their
intensity, K 1 . For N-look SAR data, Kv is 1 / N .
probability distribution is far from symmetrical, because v
they have the Rayleigh and the Negative Exponentially
probability density functions (PDF), respectively that In a moving window of 9x9, 11x11 or larger, pixels within
produce biased estimation. Consequently, to remove the these new sigma ranges are included in the application of to
bias, the sigma ranges were recomputed based on the retain edges and fine details, and then the MMSE filters are
corresponding distribution functions. The idea is to preserve again applied to the selected pixels. However, when
the mean value. For the general case of the mean value of applying the MMSE filter in the final filtering, we found
1, the sigma range of the intensity SAR data is computed. that the speckle noise standard deviation K v has to be
We have found that 90% sigma range works better than the
adjusted, because the valid range of PDF is limited by the
typical two sigma range of 95.5%. The result is shown in
sigma range. The revised K~ as shown in table 1, has a
Table 1. v
value smaller than the original K v . The data would be over
Table 1 The revised 90% sigma range ( I1 , I 2 ) for 1-look to 4-
smoothed, if the original K v were applied.
look intensity data
I1 I2 ~
K
# of looks v 3.2 Preservation of point targets
1 0.084 3.941 0,8191
2 0.221 2.744 0.5699 Preservation of signatures from point targets and manmade
3 0.313 2.320 0.4624 structures are desirable for image interpretation and other
4 0.378 2.094 0.3991 applications. These high return pixels are generally
produced by the double bounce scattering mechanism, or by
We note that, due to the asymmetrical PDFs, the interval direct specular reflection. Backscattering signatures of
( 1 I1 ), between I1 and the mean value of 1, is always point targets are significantly different from that of
distributed media. We re-emphasize that backscattered
smaller than the interval ( I 2 1 ). The new sigma interval, signatures from point targets are dominated by a small
for any a priori mean value x ~ can be easily computed from number of strong scatterers, contrasting to the speckle
~ , I x~ ). The original sigma filter
these two tables as ( I 1 x characteristics of distributed media. In other words, they do
2
~ that not possess the typical speckle characteristics.
used the center pixel in a moving window as x Consequently, they have to be processed differently from
produces the isolated dark pixel problem. The improved backscattered signatures of distributed media.
sigma filter applied the minimum mean square estimation
(MMSE) filter in the 3x3 window as estimate of the a priori The algorithm is based on the detection of strong scatterers
mean x ~ that effectively reduced the isolated dark pixels
of several pixels in size. The first step of this algorithm is
and improved overall speckle filtering effect. The minimum to compute the 98 percentile of all pixels of the SAR data to
mean square filter has the form, be filtered. We denote the 98 percentile as Z 98 . If the pixel
xˆ (1 b ) z bz (1)
to be filtered has its value greater than Z 98 , we check the
number of pixels in a 3x3 window greater than Z 98 . If
IV - 22
Authorized licensed use limited to: UNIVERSITY TECNOLOGI MARA. Downloaded on December 29, 2009 at 15:21 from IEEE Xplore. Restrictions apply.
the total number greater than a threshold Tk , we consider
all those pixels belong to the point target, and their values
will be retained. In other words, they will not be filtered.
The threshold Tk is typically set between 5 and 7 for
general applications.
5. EXPERIMENT RESULTS
IV - 23
Authorized licensed use limited to: UNIVERSITY TECNOLOGI MARA. Downloaded on December 29, 2009 at 15:21 from IEEE Xplore. Restrictions apply.
The color composition using Pauli vector is applied here Fig. 2, it clearly demonstrated the effectiveness of this
with red for |HH-VV|, green for |HV|, blue for |HH+VV|. filtering algorithm.
The original is shown in Fig.1A revealing various scattering
mechanisms: city blocks shows high returns from double 6. SPECKLE FILTERING OF DRAL-
bounce and specular scatterings, ocean areas from surface POLARIZATION SAR DATA
scattering, and park areas from volume scattering. Figure
1B shows the results of a 5x5 boxcar filter revealing The recent successfully launches of ALOS, RADARSAT-2
considerable blurring of the PolSAR image. The result and TerraSAR-X produce large volumes of high resolution
from the proposed method is shown in Fig. 1C. Compared dual-polarization SAR data. For coherently processed dual-
with the original and the 5x5 boxcar result, the proposed pol data, such as (HH and HV), (VV and VH) and (HH and
method clearly shows effective speckle reduction while VV), it is important to preserve the correlation between
retaining distinguishable details and strong point targets. polarizations just as for fully polarimetric SAR data. The
proposed algorithm can easily adopted for dual-pol case
based on the 2x2 covariance matrix. The point target
detection will be based on co-pol intensity data, and both
polarization data are to be used for filtering pixels from
distributed media.
7. CONCLUSION
REFERENCES
IV - 24
Authorized licensed use limited to: UNIVERSITY TECNOLOGI MARA. Downloaded on December 29, 2009 at 15:21 from IEEE Xplore. Restrictions apply.