Application Of Directional Filter Banks In SAR Image Processing

Synthetic Aperture Radar ( SAR ) can work under all weather, day-and-night and long distance . The application of high-resolution radar image has become a hot research t- opic in this field. Multiresolution Geometry Analysis ( MGA ) which takes Directional Filter Banks ( DFB ) as it's central part can represent the 2-D singularity signals most effectively and arose another tide after wavelet transform. MGA has been widely used in SAR image processing fields such as speckle reduction, image segmentation and ta- rget detection for it's excellent performance. Supported by the Natural Science Fund of Jiangsu Province (No.BK2001047) and Aeronautics Science Fund (No.04D52032 ), we analysed the theoretical constructure and carry out DFB with biorthogonal ladder- structure . Then we apply them into speckle denoising and target detection in SAR im- ages. Main works are as follows:1,This paper introduces the main fields of SAR image processing--- speckle denoising and target detection and provides some parameters to evaluate the quality of the SAR images.2,The directional filters is the hard-core in MGA ,we study the development of the directional filter banks and MGA . We carry out the diamond directional filter with biorthogonal ladder structures and presente the implement of Contourlet transform and Nonsubsampled Contourlet transform(NSCT).3,Combined with characters of SAR images, this paper proposes a speckle denoising method with soft thresholding algorithm based on Contourlet transform with ladder structures .A new speckle denoising algorithm based on Nonsubsampled Contourlet transform and Mean-Shift clustering algorithm was also presented. The denoised images obtained by these algorithms gain pretty good visual effect and all indices are enhanced effectively comparing to the other denoised images.4,A new method for detecting targets with NSCT energy feature is proposed based on the frequency characteristics of clutter background and target. Meanwhile, another target detection algorithm was also proposed for the difference between targets and background in NSCT field after been decomposed.