Research On Registration Techniques Forsar Images

With the development of the SAR (synthetic aperture radar) techniques, SAR playsan important role in the military affairs and civilian applications. Registration of SARimage is a fundamental task in SAR image processing used to geometrically match twoor more images, for example, at different time, from different sensors or from differentviewpoints. Registration of SAR image is becoming an active research area. AndRegistration of SAR image has been widely used in mapping, navigation, informationfusion and so on.The registration techniques of SAR image are discussed in this dissertation. Theregistration techniques have two major classes: feature-based and gray-based. First theclassic gray-based technique, correlation coefficient, is discussed. Then severalfeature-based registration techniques are presented. The main contents and importantresults of this dissertation may be summarized as follows:The registration technique based on LBP characteristic. The local characteristic isused to register SAR images.The registration technique based on line characteristic. The line characteristic ofimages is extracted with Gabor filter. And the line segments are grouped. Then theHausdorff distance is applied to estimate the parameters of transformation between twoimages.The experiment results show, the correlation coefficient technique increases thecorrection ratio, the technique based LBP is stable, which is available to the SARimages with different sizes, or different scenes, but the techniques based on CrossCorrelation and Gabor filter are suffered with the size of images.