Radar image speckle noise reduction using Kalman filtering

Student: Kelvin Doo

Supervisor: Dr. John Homer

Category: Electrical Engineering Thesis Project

Synthetic aperture radar (SAR) imagery is gaining increased usage as more systems become available and more applications are being developed including apace imagery. The popularity can be attributed to a great extent to their along track linear, resolution characteristic, which is independent of range. But an “unfortunate” problem with SAR imagery is the high level of noise, often called speckle since SAR systems rely upon coherence properties of the scattered signals. The speckle effects reduce the utility of SAR imagery a great deal. This thesis will address the causes of speckle and many of the current methods that are used to reduce the speckle noise. Kalman filtering method will then be derived and adapted for use on SAR imagery taking into account the special properties of this type of filter and images. A program called MATLAB will be used to apply the Kalman filtering method and the most commonly and easily used Intensity Summation (IS) method to SAR multi-pass images obtained by the ERS-1 satellite over Bonn in Germany. From the results obtained, we can determine which is the optimal method of speckle reduction by measuring the equivalent number of looks (ENL). It is shown that Kalman filtering method will obtain a much higher ENL which means a higher degree of speckle reduction but it is much more computationally complicated and intensive compared with using IS method which is the trade-off while using Kalman filtering.

Poster Presentation (PDF)

Thesis Document (PDF)