199 
THE EFFECT OF SPECKLE REDUCTION ON COHERENCE ESTIMATION 
AND INTERFEROMETRIC PHASE OF INSAR 
Y. L. Chen 1 *, C. Huang 1 , X. L. Ding 2 , Z. W. Li 3 , L. Zhang 2 
1 Shanghai Astronomical Observatory, CAS, Shanghai 200030, China 
2 Dep. of Land Surveying & Geo-Informatics, The Hong Kong Polytechnic University, Hong Kong 
3 School of Info-Physics and Geomatics Engineering, Central South University, Changsha 410083, Hunan, China 
ABSTRACT: 
Synthetic aperture radar (SAR) images always suffer from strongly speckle noise, which is a great obstacle to coherence (or 
correlation coefficient) estimation and interferometric phase, furthermore hinders the accuracy of phase unwrapping and information 
extraction with SAR interferometry (InSAR) technique. In this article, we propose to suppress the speckle of SAR images with 
several self-adaptive filtering methods so as to evaluate how the filtering influences the coherence and the quality of phase map by 
quantitative and qualitative analysis. As an example, ERS SAR interferometric pair covering Shanghai is used for this study. The 
filtering algorithms we adopted include enhanced Lee, enhanced Frost, Gamma Map, and Kuan filtering, etc. In order to choose the 
optimal algorithm, we used four kinds of technical criterions, such as smoothness index, speckle noise index, equivalent number of 
looks (ENL), and normalized mean, etc. By comparing the former criterions, we can determine which kind of filtering will be used in 
the following data processing. Once the appropriate filtering is determined, we thus could move on and carry on operations of 
corigistration, Phase Conjugate multiplication, and flat earth phase removal, etc. Finally, we have got the coherence and 
interferometric phase maps by using the unfiltered and filtered interferometric pairs, respectively. The mean, standard derivation 
(SD), and statistical distribution chart are used to estimate the change of coherence. The mean, SD, and sum of phase difference 
(SPD) are adopted to evaluate the quality of interferometric phase map. The results show that the speckle reduction did not improve 
the quality of interferometric phase map and nearly have no impact on the coherence coefficients. The next study is which kind of 
other measures should be adopted under the condition of small temporal and spatial baseline in the interferometry. 
1. INTRODUCTION 
InSAR (Interferometric Synthetic Aperture Radar) is a new hot 
space technique in Microwave Remote Sensing as well as 
Geodesy for its great ability to extract landscape topography 
and its deformation during day and night under any weather 
conditions by involving two or more SAR images, see e.g. 
Zebker[l], Ferretti [2], Hassen[3], for a general introduction of 
this technique. 
Speckle in SAR image is a troublesome factor, which may has a 
great impact on the accuracy of terrain, land-use classification, 
and scene interpretation. For the sake of extracting more 
accurate and useful information, speckle filtering become very 
important in the case studies. 
From the open issued literature, we can find that researches on 
SAR speckle filtering, InSAR data processing and its 
applications have got enough concerned during the past decades. 
However, the studies on how the speckle filtering influenced the 
coherence estimation and interferometric phase seems to be paid 
little attention, which is an unignorable factor according to our 
preliminary analysis. In this paper we emphasized our focus on 
speckle filtering technique and attempted to investigate the 
relation between the SAR speckle filtering in the time domain, 
and the coherence and interferogram. 
2. SAR SPECKLE FILTERING METHODS 
many kinds of filtering methods based on statistical and non- 
statistical models, such as (enhanced -) Lee, (enhanced -) Frost, 
and Kuan and Gamma MAP filtering, mean filtering, median 
filtering and order statistical filtering, etc. In frequency domain 
the wavelet transform is used most frequently. 
Filtering algorithms based on statistical model have the 
advantage of adjusting the weights self-adaptively, it is thus 
widely used in relatively complicated signal processing e.g. two 
dimensional image filtering. We proposed to use four filtering 
methods, such as enhanced Lee, enhanced Frost, Gamma Map, 
and Kuan filtering, and then choose the most congruent one to 
continue our research that is how the speckle filtering influence 
the interferometry. It equals to investigating the effect of SAR 
pre-filtering in the InSAR technique. 
3. DEFINITION OF COHERENCE ESTIMATOR AND 
INTERFEROMETRIC PHASE 
Coherence coefficients and the quality of interferometric phase 
are usually used to evaluate the interferometric quality. SAR 
phase interferogram and coherence can be generated by 
complex conjugate multiplication of two coregistered SAR SLC 
images. 
The coherence (or correlation) coefficients generally can be 
defined as 
In past years multilooking operation is generally used to reduce 
the speckle of the SAR images. However, for InSAR technique, 
this operation will breach the premise of phase preserving. 
Therefore this method is out of our consideration in this paper. 
Besides multilooking, the filtering can be carried on in time 
domain and frequency domain. In time domain, there appeared 
\f\ = 
M,N 
X s \(Uj)s\{iJ) 
'=1,7=1 
M,N 
X s ^j)s\(iJ) X s 2 (i,j)s* 2 (i,j) 
1=1,7=1 '=1,7=1 
(1)