Published in IET Radar, Sonar and Navigation
Received on 7th June 2013
Revised on 31st October 2013
Accepted on 16th November 2013
doi: 10.1049/iet-rsn.2013.0202
ISSN 1751-8784
Improving the accuracy of local frequency estimation
for interferometric synthetic aperture radar
interferogram noise filtering considering large
coregistration errors
Jiao Guo
1,2
, Weitao Zhang
2
, Yanyang Liu
2
, Longsheng Fu
1
1
College of Mechanical and Electronic Engineering, Northwest Agriculture and Forestry University, Yangling, Shanxi
712100, People’s Republic of China
2
National Key Lab of Radar Signal Processing, Xidian University, Xi’an 710071, People’s Republic of China
E-mail: jiao.g@163.com
Abstract: This study deals with the problem of estimating the local frequencies for interferometric synthetic aperture radar
(InSAR) phase image (i.e. interferogram) noise filtering considering large coregistration errors. The estimation of local
frequencies is frequently applied to achieving high performance of interferometric phase noise filtering, which is a key step in
InSAR processing procedures. Unfortunately, the generated interferograms suffer seriously from coregistration errors
especially for complicated topographies, thus imposing strong limits on the accuracy of local frequencies estimation by the
existing methods. Taking large coregistration errors into account, the proposed method makes full use of the neighbouring
pixels to construct joint pixel vector and take the separation extent of the signal and noise subspaces as the criterion to
accurately estimate the local frequencies so that the effects of the envelope misalignment can be mitigated. Theoretical
analysis and the simulated data as well as the real ERS-1/2 data show that the presented method has the ability to provide
accurate estimation of local frequencies even if the coregistration error reaches one pixel.
1 Introduction
Synthetic aperture radar interferometry (InSAR) is one of the
most important remote-sensing techniques to acquire terrain
digital elevation models (DEMs) [1, 2]. InSAR illuminates
the same ground scene to obtain two or more complex SAR
images from slightly different incident angles. Also, the
phases of the SAR image pair are processed in
interferometric way to determine the height of each
scattering element in the observed scene (i.e. each pixel of
SAR image). This advanced remote-sensing technique,
which can work either day or night and through cloudy
cover, has numerous applications in the fields of
topography, geomorphology, seismology, and so on [1–3].
As is well known, interferometric phase noise suppression
is a key step in InSAR processing procedures. To achieve fine
performance of phase noise suppression, it is of extreme
importance to obtain more filtering samples that are
independent and identically distributed (i.i.d.) from the
adjacent pixels. The i.i.d. condition requires that the
adjacent pixels must have an identical terrain height. In
other words, the height difference of the adjacent pixels
inevitably results in poor performance of phase noise
suppression, consequently deteriorating the DEM accuracy.
Therefore, the effects of neighbouring terrain changes must
be removed especially for complicated topography, which is
usually called topography compensation. Of course, if a
prior coarse DEM of the observed scene is available, the
effect of terrain changes on the interferograms can be
removed conveniently. However, in practice the processor is
not always provided with a coarse DEM available.
Consequently, the parameter estimation techniques are
usually employed to estimate local frequencies of
interferograms for topography compensation. On the one
hand, the estimation of local frequencies can be completed
by the existing methods such as the maximum likelihood
(ML) method [4] or the multiple-signal classification
(MUSIC) method [5–7]. However, the effecti veness of
these methods strongly depends on the coregistration
accuracy of each pixel in the SAR image pair, that is, these
methods may fail in the presence of large coregistration
errors. Unfortunately, the accuracy of coregistration
methods is always finite, especially in the areas with
complicated topography, since the coregistration offset is a
function of terrain changes. For example, if the
coregistration error reaches one pixel, the interferograms are
decorrelated completely and seem to be purely noisy, and
the methods mentioned above inevitably fail to correctly
acquire local frequency estimates. On the other hand, the
proposed method in the referee [8] uses joint subspace
projection method by constructing joint pixel vector.
However, it is still necessary to select the neighbouring
www.ietdl.org
676
&
The Institution of Engineering and Technology 2014
IET Radar Sonar Navig., 2014, Vol. 8, Iss. 6, pp. 676–684
doi: 10.1049/iet-rsn.2013.0202