Development and analysis of wavelet based speckle reduction algorithms for synthetic aperture radar images

Abstract

There has been a growing interest in Synthetic Aperture Radar (SAR) imaging due to its ability to operate under varied weather conditions. SAR is also capable of penetrating through cloud and soil. SAR systems have its importance in a variety of applications such as remote sensing for mapping search-and-scene rescue, mine detection, and target recognition. In this thesis, Dual Tree Complex Wavelet Transform(DTCWT) based despeckling algorithm is proposed for SAR images, considering the significant dependencies of the coefficients across different scales. In this thesis, a method is proposed to reduce the number of false edges in the MRGoA detector output. This algorithm discriminates the objects and the false edges automatically by applying the principle of component connectivity and Shannon s entropy. Further, an automatic threshold selection method is also proposed for the existing Ratio of Exponential Weighted Averages (ROEWA) based edge detector. Otsu s nonparatmetric and unsupervised principle is used for automatic threshold selection in the ROEWA edge detector. In the proposed method, optimal threshold is selected by maximizing the seperability of the classes in gray level by incorporating a simple search strategy. From the experimental results, it is observed that the automatic threshold selection methods proposed for MRGoA and ROEWA has better edge detection capability. The Pratt s Figure of Merit (FOM) performance of the edge detection algorithms is improved when speckle reduction is done prior to edge detection. FOM value of the proposed entropy based MRGoA is increased by 5% when compared to the existing MRGoA edge detector. Also, edge detection by the automatic threshold selection based ROEWA is more efficient than ROEWA with predefined thresholds. newline newline newline