. Istanbul 2004 
THE CONTRAST RESEARCH OF THE METHODS OF RESTRAINING THE SPECKLE 
NOISE OF SAR IMAGES 
Zhiyong Wang *^ 
lixian Zhang? Tongxiao Wang? 
“ Shandong University of Science and Technology, Tai'an, China — (wWzywlp,lex6088)@ 163.com 
b r3. ^ 5 . . n . — . 
Chinese Academy of Surveying and Mapping, Beijing, China —stecsm@public.bta.net.cn 
KEY WORDS: SAR, High resolution, Method, Analysis, Model 
ABSTRACT: 
There commonly exists speckle noise in the SAR images. It badly disturbs the extraction and interpretation of the information of the 
objects, influences the application of the SAR images. So it has important meaning to reduce the speckle noise in the SAR images. In 
this paper, the mechanism and mathematical model of the speckle noise is firstly analyzed. Then the methods of filtering speckle noise 
is introduced, especially several local region adaptive filters. Two filter models are summarized to direct the construction of the new 
filter. Then for the typical filter, the contrast tests are carried out. At the last, it brings out a quantitative evaluation index to direct the 
filtering the speckle noise and the construction of the new filter. 
1. Introduce 
Because Synthetic Aperture Radar (SAR) can pass the clouds 
and mist and has the all-weather and full work ability, it has 
unique superiority with compare with the visible and infrared 
remote sensing. Therefore, it has been shown more and more 
bigger application latent capacity in many domains, such as 
surveying and mapping, geology, hydrology, ocean, ecology, 
environmental monitoring, glacier charting and military affairs. 
Synthetic Aperture Radar (SAR) images are generally corrupted 
by speckle noise. The presence of the speckle noise in the SAR 
images is the major obstacle to interpret, classify and analyze 
SAR images for different application purposes of remote 
sensing users. The speckle noise badly disturbs the extraction 
and interpretation of the information of the objects, influences 
the application of the SAR images. Therefore many speckle 
noise filters have been proposed for several decades. 
2. The mechanism giving rise to speckle noise and 
mathematical model of the speckle noise 
Speckle noise is a physical effect, which occurs when coherent 
light is reflected from an optically rough surface. 
The characteristic speckle effect of radar images results from 
the destructive and constructive interference among the echoes 
of individual surface scatters within a resolution cell. Therefore 
129 
the resultant pixel can differ extremely from its average grey 
value. These grey value variations between adjacent pixels lead 
to the typical salt and pepper appearance of SAR images. 
The multiplicative speckle model supposes the speckle to be 
fully developed : Within each resolution cell there must be a 
large number of scatterers whose phase and amplitude are 
statistically independence ;the different scatterer amplitudes 
must belong to the same statistical distribution ;their phases are 
[0,211] uniformly distributed . 
When the ground surface rough degree surpasses the radar 
wavelength (for example, RadarSAT is 5.6cm and ERS-1/2 is 
5.6666cm), and there are many scatters within a resolution cell, 
the speckle is regarded as fully developed. An appropriate 
model for fully developed speckle is a multiplicative fading 
random process F (Ulaby et al .1986 a ) 
H(x, y) * R(x.v)* F(x. y) (1) 
where (x,y) are the spatial azimuth and slant range coordinates 
of the resolution cell center , I is the observed intensity 
(speckled observed radiance ), R is the random radar reflectivity 
process (unspeckled radiance). F is a second order stationary 
random process, statistically independent of R, with unit mean 
r 2... 
M ,, — 1) and whose variance (V. — O; ) is inversely 
f f f