Speckle Noise Reduction in SAR Images

Student: Wei Lin Jaclyn Chang

Supervisor: Dr. John Homer

Category: Electrical Engineering Thesis Project

In this thesis, two methods of speckle noise reduction in synthetic aperture radar (SAR) images are investigated. They are the Intensity Summation method and the soft-thresholding method.

These two methods were implemented and tested using the software Matlab. Simulated images generated using Matlab and a real SAR image obtained form the ERS-1 satellite over Bonn in Germany, were utilized to test the noise reduction techniques. There are two approaches used to apply the noise reduction techniques and they are the segmentation and the moving segment approach. With the results obtained, we measure the effective number of looks (ENL) and mean square error of the de-noised images to determine the better noise reduction technique

It was concluded that the de-noised image by moving segment gave a higher quality de-noised image than the segmentation approach. Given the results generated from the Intensity Summation and Soft-thresholding noise reduction techniques, de-noising by soft-thresholding is found to be more accurate than the intensity summation method. The mean square error formula and the cross-sectional plot of the de-noised image indicate that the Intensity Summation method does not retain image structure well due to the tendency of flattening it. Soft-thresholding unlike intensity summation does not flatten image structure but retains it. This was confirmed from the cross-sectional view of the de-noised images. Both methods reduce speckle noie and yield high ENL values but the soft-thresholding method gives rise to a more accurate and desirable outcome than the Intensity Summation method

Poster Presentation (PDF)

Thesis Document (PDF)

Other Related Files