117 
WAVELET SPECKLE REDUCTION FOR SAR IMAGERY BASED ON EDGE 
DETECTION 
Yingdan Wu a ' , Xiuxiao Yuan 3 
a School of Remote Sensing and Information Engineering, Wuhan University, 129 Luoyu Road, Wuhan 430079, China 
Wudan 1982@hotmail.com 
Commission I , WG 1/2 
KEY WORDS: Image processing, Filtering, Feature detection, Geography, Image understanding 
ABSTRACT: 
This paper introduces a wavelet transform speckle reduction algorithm for Synthetic Aperture Radar (SAR) imagery based on edge 
detection. Existing speckle algorithm can efficiently reduce the speckle effect but unfortunately also, to some degree, smear edges and 
blur images. In this paper, the original image is firstly logarithmic transformed and decomposed with multi-scale wavelet transform. 
For each pyramid level, edge cross points are detected by using the wavelet transform modulus maximum. The multi-scale and edge 
fusion strategy enables to detect only edge crossings and ignore the local oscillations. Then local wavelet soft-threshold filter is applied 
to the area that there is no edge crossing point. Repeat it through the image pyramid levels, and the despeckled image is reconstructed. 
The experiments have been carried out to verify the method proposed in this paper and the results is elaborately analyzed. The results 
have shown that our method can not only filter the speckle noise efficiently, but well preserve the image detail in the SAR imagery 
1. INTRODUCTION 
Synthetic Aperture Radar (SAR) is a kind of high resolution 
imaging system. It generates imagery which does not depend 
on time and weather conditions. It has the ability to penetrate 
through some depth of the soil or vegetation. SAR imagery is 
used in many fields, such as agriculture, forestry, geology, 
hydrology and so on (Fetter et al., 1994). Due to the coherent 
nature of the imaging system, it is inevitably that speckle exists. 
The presence of speckle reduces the radiometric resolution of 
the image and the detectability of the image feature. It is 
usually desirable to reduce the speckle noise prior to image 
applications, and speckle reduction is becoming a commonly 
used routine process. 
Speckle in SAR imagery is multiplicative noise. As a 
consequence, a number of filtering algorithms dealing with 
multiplicative noise have been proposed. The most notable 
include the Lee (Lee, 1980), Kuan (Kuan, 1987), and Frost 
(Frost, 1982) filters. These filters, aims at minimizing the mean 
square error (MSE), are derived from the speckle model, i.e., 
assuming speckle is a multiplicative noise random variable, 
with mean of one. By examining the derived formulas, however, 
the Lee and Kuan filters can be considered as adaptive-mean 
filters, and the Frost filter is an adaptive-weighted-mean filter. 
Meanwhile other filters not derived from speckle models, such 
as the mean filter, median filter, geometric filter (Crimmins, 
1986), and wavelet transform filter (Dong et al, 1998) have also 
been applied for SAR speckle reduction. Compared with the 
traditional statistical speckle filter, wavelet transform filter 
have several characteristics: (i) they preserve high frequency 
information; (ii) the balance between speckle reduction and 
detail preservation can be adjusted; (iii) they require no 
knowledge of the standard deviation of speckle. 
Existing speckle filtering algorithms can efficiently reduce the 
speckle level. However, these algorithms also, to some degree, 
smear edges and blur images. Smoothing uniform areas while 
preserving and/or enhancing edges is difficult to accomplish. In 
the frequency domain, the former requires abandonment of 
high frequency components while the later needs the 
preservation of high frequency components. Adaptive filters 
take account of speckle distribution models and compute local 
statistics within a moving window and assign new values 
accordingly, leading to better results. 
This paper introduces a new algorithm, which incorporates the 
wavelet transform filter and edge detection altogether, to 
achieve the goal of smoothing uniform areas and preserving the 
edges. The experiments have been carried out to verify the 
method proposed in this paper and the results is elaborately 
analyzed. Although the derivation of the algorithm is not based 
on the speckle model, by applying logarithmic transform to the 
original image, the algorithm is also applicable to the 
multiplicative speckle filtering. 
2. OVERVIEW OF ALGORITHM 
A wavelet transform speckle reduction algorithm based on edge 
detection is proposed in this paper. It implements as follows: 
with consideration of the particularity of speckle noise, firstly 
we apply logarithmic operation on SAR imagery to convert 
multiplicative noise into the additive noise model. And 
decompose the image into several levels by wavelet transform. 
In each level, before the filtering, the edge information is 
acquired. Firstly, the edge information from the higher level is 
projected to the current level. Then Wavelet Transform 
Modulus-Maximum Algorithm is applied to detect the 
candidate edge points in current level. By fusing them, the final 
edge information in this level is obtained,