Implementation Of Contourlet Transform And Application In SAR Image Processing

Contourlet transform is a new Multiresolution Geometry Analysis(MGA) method which can represent images most effectively. Especially it can represent the 2-D singularity signals effectively. It becomes a hot topic of researches by the scholars inside and outside our country. Synthetic Aperture Radar (SAR) is a kind of high resolution imaging radar, which has been developed since World WarⅡ. High-resolution radar image can be obtained under all weather, day-and-night and long distance conditions. Now Contourlet transform has more and more applications in SAR image processing such as speckle denoising because of its excellent performance. Supported by the Foundation Research Project of Jiangsu Province (The Natural Science Fund N0.BK2001047)"Research on SAR Image Evaluation", we study the theory of Contourlet transform detailedly and carry it out in this thesis. And then we apply the Contourlet transform to the speckle denoising in SAR image.1,This paper introduces the main fields of SAR image processing—speckle denoising and provide some parameters to evaluate the quality of the images; This paper also introduces MGA and emphasizes on the Contourlet transform.2,The directional filter is very important in MGA especially in the Contourlet transform. In this paper, we do some preparatory study for the implementation of Contourlet transform——the diamond filters and the fan filters.3,This paper analyses the theory and the structure of the Contourlet transform and carries the transform out with software based on the binary tree structure which is made up of the diamond filters and the fan filters realized above.4,Combined with characters of SAR images and the relativity of the coefficients, this paper proposes a new speckle denoising algorithm for SAR image based on Contourlet transform. This new algorithm has a foundation of the old denoising algorithms before. The denoised images obtained by these algorithms gain pretty good visual effect and all indices are enhanced effectively comparing to the other denoised images.