Registration Of InSAR Image And Parallel Algorithm Study

Interferometric Synthetic Aperture Radar (InSAR), as a rising technology in deriving elevation data and slight deformation of earth surface, has drawn extensive attention in various fields such as topographic mapping, monitoring of earthquakes, volcanoes, land subsidence and glacier dynamics and other thematic mapping applications. Registration of two or more images of the same scene is an important procedure in InSAR image processing that seeks to extract differential phase information not obtainable from each one of these images. Meanwhile, the accuracy of this step is crucial to the reliability of subsequent image processing and final results of the data processing chain, meeting subpixel precision request. After analyzing systematically some conventional InSAR registration methods, his paper presents two multi-step image matching algorithms based on the characteristics of InSAR image. One approach is integrating correlation-registration and Max-spectrum image registration, the other is integrating correlation-registration and least square-registration. In coarse mage registration, two approaches all utilize correlation-registration to obtain pixel-accuracy registration results, offering crude control points pairs for the next matching. In precise image registration, two approaches respectively apply Max-spectrum and least square-registration method to register images, the matching accuracy reaches to about subpixel.For purpose of proving validity of the registration method, Experimentations are implemented on several different test sites and meet the request of registration. The paper compares further the two algorithms from some aspects of validity, reliability, registration quality and computing time and obtain some beneficial results.So far, parallel processing is a perfect alternative for InSAR image processing for its capability of fully utilizing available computing resources and high scalability. Considering the requirements of processing highly massive image data in efficiency, accuracy and computation cost, the robust and reliable computation architecture is constrcucted on cluster of PCs by MPI middleware .As a trial on parallel processing, the paper implements a parallized registration algorithm on cluster of PCs and parallel computing environment , showing advantages compared with serial coregistration algorithm. By a few performance criterion tests, several important factors that have greatly effects on parallel system have been analyzed in detail.