SAR Pixelwise Registration via Multiscale Coherent Point Drift With Iterative Residual Map Minimization

Abstract

Due to the severe speckle noise and complex local deformation in synthetic aperture radar (SAR) images, robust pixelwise registration with high accuracy is an important problem but is far from being resolved. The core of this problem is how to establish a precise deformation field that maps every pixel to its corresponding pixel with high accuracy. To address this problem, a novel SAR dense-matching algorithm, which includes high-accuracy landmark generation and a precise deformation field parameter estimation, is proposed in this article. First, a strategy for generating enough well-distributed landmarks is proposed by designing patch matching of improved scale-invariant feature transform features based on phase correlation and the gradient method. Furthermore, a multiscale coherent point drift (MCPD), powered by iterative residual map minimization, is designed to reliably match landmarks and estimate precise field parameters. Both simulated deformed SAR images and real SAR images are utilized to evaluate the performance of the proposed method, and the experimental results demonstrate that the proposed method provides better registration performance than previous methods in terms of both accuracy and robustness.