Research On Super-resolution Reconstruction In SAR Image

As the representative of the microwave remote sensing technology, SAR has been widely used in earth science and some other relevant fields because of its unique earth observation ability. As the particularity of SAR imagery will inevitably cause the speckle noise in images, even if scattering coefficients of radar irradiation target are very even, the image formed by echo will not be even in theory. It's necessary to do a preprocessing for the low SAR images that contained lots of noise so that the image preprocessed will be smoother and a large number of spots noise will be suppressed. So far, the technology that suppresses coherent spots has not been ideal in application. Therefore, based on what has been done by predecessors, this paper is devoted to research on the problems and effects of removing the coherent spots by the method of PDE and the method of Contourlet Transformation. What has mainly been done in this paper are as follows:(1) The principle of SAR imagery and the feature of coherent spot noise have been analyzed. A brief introduction to spatial filtering technology and frequency domain filtering of the SAR image has been made. Moreover, both of their advantages and disadvantages has been analyzed as well.(2) Based on the low resolution of SAR images, an approach combining P_M nonlinear diffusion equation and edge enhancing diffusion equation was proposed, which used P_M nonlinear diffusion equation's advantages that remove the noise and protect images'edges at the same time. The approach proves edge enhancing diffusion equation's feature that can enhance the edge of images, avoid the disadvantage that P_M nonlinear diffusion equation can't improve the edge of image and edge enhancing diffusion equation may produce the line-like structure. The experimental results show the effectiveness of this method. (3) A method is proposed that makes nonlinear diffusion in Contourlet domain. This method makes nonlinear diffusion for the decomposition coefficient in Contourlet domain and control the strength and direction of diffusion by using the diffusion coefficiency constructed by amplitude values. By comparing the wavelet threshold method with the traditional Contourlet threshold method, the experiment results show that this method is more effective to remove the noise and creates satisfying visual continuity.