基于压缩感知的遥感影像弹性配准方法

针对遥感影像由于载荷类型、观测角度、地形起伏等内外部因素造成的影像局部几何畸变,而基于全局配准方法制约着影像配准精度的提高,基于像元的弹性配准方法可大幅提升遥感影像的配准精度,但是存在运算效率这一瓶颈等问题,该文利用像元弹性配准参数的稀疏性,提出一种基于压缩感知的弹性配准方法。通过对遥感影像像元梯度幅值响应较强的点进行随机抽样,形成观测样本点集,采用弹性配准局部参数解算模型求解样本点平移参数;通过压缩感知稀疏重构算法重构影像各像元平移参数。实验表明,在配准精度差异较小和一定的参数设置条件下,该方法可显著提高弹性配准运算速度。

Study on remote sensing image elastic registration algorithm based on compressive sensing

The algorithms based on global rigid model cannotresolve problems of local geometric distortion caused by the internal and external factors such as different remote sensing payloads,observation angle and times,topography etc.The pixel-based elastic registration method in some literatures can significantly improve remote sensing image registration accuracy,but the operation efficiency becomes the bottleneck of the technology applications.In this paper,a new elastic registrationmethod based on compressive sensingwas proposed according to the sparity oflocal registration parameters.The method obtains the sample points which have stronger gradient magnitudeby random sampling,and calculates the elastic registration local translation parametersof the sample points.Finally,each pixel's local translation parameters of the image can be obtained by compressive sensing sparse reconstructionalgorithm.Experiments show that this method can significantly increase the elastic registration processing speed under smaller differences in registration precision and a certain set of parameters conditions.