Coregistration based on stochastic parallel gradient descent algorithm for SAR interferometry

Abstract

The coregistration of complex image pairs is a very important step in interferometric synthetic aperture radar (InSAR) data processing. This article proposes a coregistration method based on the stochastic parallel gradient descent (SPGD) algorithm. Stochastic parallel perturbations are imposed on the translation coefficients of the polynomial coregistration model to make the performance evaluation function converge to a global extremum, which allows the translation coefficients to be obtained, and then the coregistration is achieved after resampling. Data processing of images from Kashgar and Mount Etna show that the proposed method is effective and robust. Furthermore, a series of experiments is designed to evaluate the convergence characteristics of the proposed method, which indicates that it has a stable convergence process and good robustness.