Research On PDE Based Speckle Reducing Methods In SAR Image

Nowadays, Synthetic Aperture Radar becomes a most important resource of remote sensing information. It has been widely applied in military and civil aspects. But the images of SAR always contain a lot of speckle, which is a kind of multiplicative noises, degrade the radiant resolution and SNR of the image seriously. Speckle also makes it difficult to interpret and classify the images of SAR. So reducing speckle, which is one of the most challenging problems in the fields of image recovering and enhancing, becomes a very important and attractive research topic recently. But until now, it is still far from satisfactorily solved.There are about four kinds of methods to reduce speckle. The first is N-looks processing, which average the multi-looks of the same region. The second is adaptive filters based on the statistics of the SAR image. The third is using wavelet transform to remove the speckle. The fourth is anisotropic diffusion and total variational method based on PDE.In this paper a comprehensive classification and overview of these speckle reducing techniques is given, further discussion and investigation focused on PDE based methods are also carried out. On the basis of an in-depth analysis, we propose a novel class of practicable measures, a regularized speckle reducing anisotropic diffusion equation, which we believe will be proved to be useful and efficient for SAR image recovering.The innovations in this dissertation can be listed as follows:1. We gave a summary of speckle reducing methods especially focused on PDE based algorithms, and present the relationship between the conventional adaptive filters and PDE.2. We dissected the speckle reducing anisotropic diffusion method, we have found some irrationalities in the diffusion control function and proposed a new onewhich would lead to more stable resolution with edge-sensitive effectiveness. 3. We proposed a regularized speckle reducing anisotropic diffusion model.