图像配准是实现干涉合成孔径雷达（InSAR）高精度相位提取及地形高程反演的关键，大场景图像的高效高精度配准成为近年高分宽幅InSAR成像应用研究的难点问题之一。由于大场景图像中不同区域偏移量及变化规律差异较大，传统最大相干系数配准方法需多分块及插值处理，面临计算量大且配准精度低等问题。针对此问题，本文提出一种基于DFT模型的大场景InSAR高效高精度图像配准算法。该方法利用最小均方差准则构建InSAR复图像配准的DFT模型，采用四叉树自适应分块及矩阵相乘DFT快速重采样配准方法，实现大场景InSAR图像各子块区域的高效高精度亚像素配准。仿真和实测数据验证本文算法的有效性，结果表明该算法不仅可实现大场景InSAR复图像亚像素级配准，还具有较高的运算效率，其运算效率相对于传统FFT配准方法通常可提升3倍以上。

Image registration is key to high-resolution phase extraction and height inversion for interferometry synthetic aperture radar (InSAR). Recently, the registration method with high accuracy and efficiency has been one of the widely discussed issues in High-Resolution Wide-Swath (HRWS) InSAR applications. Different regions of a large scene will cause the pixel offsets of InSAR images to change dramatically. Traditional algorithms of InSAR image registration are based on maximum coherence require substantial block and interpolation processing, which may suffer from huge computational complexity and low precision.
An efficient image registration algorithm for InSAR large scenes via DFT model is proposed in this study. In the scheme, a DFT model of InSAR complex image registration is constructed based on the minimum mean square error criteria. Then, the efficient sub-pixel registration for InSAR complex images is achieved via quadtree block and matrix multiplication DFT registration.
Simulation and experimental results are presented to confirm the effectiveness of the algorithm.
Results demonstrate that the algorithm can achieve subpixel image registration of InSAR large scenes and has high computational efficiency, usually more than thrice that of traditional FFT-based registration methods.