AN INSITU SINGLE-POINTED WAVELET-BASED METHOD FOR NOISE REDUCTION 
IN SAR IMAGES 
Huan Gu a ’ *, Guo Zhang 3 , Jun Yan 3 
a State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, 
Wuhan University, Wuhan, China, 430079-guhuanl 14031@yahoo.com.cn, zhangguo_wtusm@163.com, 
yanjun_pla@263.net 
KEY WORDS: Synthetic Aperture Radar (SAR), Speckle noise, Insitu Single-pointed Wavelet-based (ISPW) Method, Visual 
quality, Statistical values 
ABSTRACT: 
Speckle noise caused by interference of electromagnetic waves complicates Synthetic aperture radar (SAR) images to be interpreted 
and further practical application, therefore reducing speckle noise and preserving more detailed information are indispensable tasks 
in pre-processing. In this paper, a new insitu single-pointed (ISPW) method based on wavelet decomposition to several scales to 
suppress speckle noise was introduced. On the basis that pixels in log-transformed images are mutually independent, we operate 
insitu operations on wavelet coefficients in both approximate and detailed parts in each scale according to distribution of 
approximate component in each wavelet decomposition scale, while not affecting others’ in order to remove more speckle noise as 
well as preserve more original edge information. Several evaluation criteria are applied to examine the performance of this ISPW 
method. Results of experiments show that ISPW method can give a better visual quality and obtain a higher signal-to-noise ratio 
(SNR), peak signal-to-noise ratio (PSNR) and edge preserving ability (EPA) as well. 
1. INTRODUCTION 
Synthetic aperture radar (SAR) plays an increasingly important 
role in offering information of earth’s surface and they are now 
widely used in various implementations such as resource 
monitoring, navigation and positioning and military command. 
However, the coherent integration brings speckle noise into 
SAR images during the process of receiving signals from 
targets towards which space borne or airborne platform sends 
electromagnetic waves (Goodman, 1976); and the introduction 
of speckle noise makes it difficult for people to interpret SAR 
images to practical purpose. Hence, it is critical to remove 
speckle noise from SAR images in pre-processing steps. 
Such many spatial-domain filters have been proposed for 
suppressing speckle noise as Lee, Frost and so forth (Lee, 1980; 
Frost, 1982), however these approaches are greatly dependent 
on size and orientation of their local windows. Usage of 
neighbourhood pixel values weakens the object signals and 
detailed edge information, the focus of surveying, mapping and 
orientation. Recently, multi-scale wavelet transform and 
threshold techniques are developed in removing speckle noise 
from image signals. According to characteristic of wavelet 
coefficient, many authors suggested various methods to select 
effective threshold, and universal threshold (Donoho and 
Johnstone, 1994), GCV threshold( Jansen, Malfait and Bultheel, 
1997) and BayesShrink threshold (Chang, Yu and Vetterli, 
2000) are three typical thresholds. It has been shown that 
compared with spatial-domain filters, wavelet threshold can 
operate a better performance in reducing speckle noise (Gagnon 
and Jouan, 1997). However, based on statistical estimation and 
global probability, these wavelet threshold methods can still not 
provide a better result in preserving more edge information. 
On the basis of these discussions, it is necessary to propose a 
new approach to suppress speckle noise in its own position 
while not affecting neighbourhood pixel information to reduce 
the loss of edge. In this paper, a new insitu single-pointed 
wavelet-based (ISPW) method, that employs distribution of 
approximate part of wavelet coefficients to locally operate on 
both approximate and detailed components, is proposed for 
denoising speckle noise in SAR images. This approach 
eliminates speckle noise locally while not involving their 
neighbourhood information. The performance of proposed 
ISPW method using several statistical values is compared to 
that of some of the existing techniques. 
2. PROPOSED METHOD FOR DENOISING SAR 
IMAGES 
2.1 Basis of proposed method 
Let y(k, 1) denote the intensity of (k, l)th pixel in a SAR image, 
s(k, 1) the noise-free image pixel which we wish to recover from 
y(k, 1), and n(k, 1) the multiplicative speckle component. 
Assuming the speckle to be fully developed, y(k, 1) is expressed 
as (M. I. H. Bhuiyan, M. O. Ahmad and M. N. S. Swamy, 
2007): 
y(k, 1)= s(k, 1)* n(k, 1) (1) 
With the log-transformation, (1) becomes 
Y(k,l)=X(k,l)+N(k,l) (2) 
Corresponding author.