Image Registration of TerraSAR-X Data using Different Information Measures

Image registration is known as an important part of generating Digital Elevation Models (DEM) with Interferometric SAR (InSAR) processing and is one of the critical preprocessing steps in remote sensing. It is used in the formation of 3-D models based on 2-D images taken at different view points as well as for mosaicking applications. The paper gives an overview of the information measures which can be used for automatic generation of reference points und finding the correspondences in the second image. The approach is validated with simulated and real image pairs originate from the TerraSAR-X satellite. Additionally to the classic correlation based methods, we will employ, compare and combine theoretical methods, based on information theory, such as mutual information, alignability in conjunction with automated bin size determination to achieve optimally. Considering the co-registration of single look complex SAR data, where correlation approaches are standard, the exploitation of similarity measures and transformation, based on information theory, is rather novel. Comparing these novel approaches with standard references with respect to quality and quantitative measures, will broaden the scientific understanding, and, furthermore, provide new insights concerning co-registration techniques for interferometric applications.