ure omps 2010
1
 
UREOMPS2010 1
 
     
 
Home
Staff
Teams
IU Teams
KU Teams
Ghana Team
Washington Team
Schedule
KU Lecture Schedule
Photos
Travel Guide
Resources
 
National Science Foundation
CReSIS
Office of Naval Research
ECSU
CERSER
 
This site is co-funded by the ASSURE program of the Department of Defense in partnership with the National Science Foundation REU Site program under Grant No. ANT-0944255
NSF DOD
 
 
 
 
 
 
Colin Axel
b

Abstract

Removal of Speckle Noise from Synthetic Aperture Radar Images

Colin Axel and Dr. Swapan Chakrabarti, Center for Remote Sensing of Ice Sheets, the University of Kansas, Lawrence, KS 66045

CReSIS flies airplanes mounted with radars in order to image ice sheets and their internal layers. To keep the airplanes as aerodynamic as possible, the antennas are restricted in size. Too small of an antenna can result in a lack of resolution, but high resolution is needed to distinguish internal layers and features of the ice sheet. To overcome this, synthetic aperture radar (SAR) is used. Using SAR, a high resolution image can be obtained by signal processing multiple images of the same target taken from different points. However, due to the return signal coming from several different scattering points, interference occurs and produces a granular, multiplicative noise known as speckle noise. This speckle noise degrades image quality and can make internal layers and features of the ice sheet indistinguishable. There are several existing methods to remove speckle noise. This investigation implements, tests, and compares four existing methods as well as one new method on real CReSIS data. The image used is an SAR image from CReSIS that originally had very little noise. Speckle noise was added using a built-in noise function in MATLAB©. The methods include: the Frost Filter, the Lee Filter, the Wavelet method, Speckle Reducing Anisotropic Diffusion (SRAD), and the Artificial Neural Network (ANN).The performance of these methods was evaluated using mean square error, root mean square error, signal-to-noise ratio, and peak signal-to-noise ratio. A plot of how each method performs given different input conditions is also included. SRAD turned out to be the best performing method in all cases.

       
b