利用光流配准进行嫦娥一号CCD多视影像超分辨率重建
-1 CCD Multi-view Images
-
摘要: 目的 月面起伏较大导致嫦娥一号CCD多视影像之间的形变是非刚性的,传统的基于仿射变换配准的超分辨率重建方法很难取得理想的效果。本文提出了一种新的基于光流配准的超分辨率重建方法,该方法采用双边总变分(bilateral total variation,BTV)超分辨率重建方法迭代求解高分辨率影像,迭代初始值由光流配准后的多视影像共同产生,并在迭代求解高分辨率影像的过程中更新光流场。实验结果表明,本文方法解决了嫦娥一号CCD多视影像超分辨率重建中的亚像素精确配准问题,重建影像的空间分辨率显著提高,细节分辨能力增强。Abstract: Objective As the deformation among multi-view images from the Chang’E-1CCD is non-rigid,giventhe non-planarity of the moon surface,the traditional methods based on the affine transform model arenot suitable for super-resolution correction.For this reason,a new optical flow based super-resolutionframework is proposed.In the framework,the BTV method is employed to estimate a high resolutionimage iteratively.The initial values for a high resolution image are generated by all multi-view imagesthat are registered by optical flow,while the optical flow fields are updated each iteration.Experimen-tal results show that the proposed method performs better for the super-resolution of Chang’E-1CCDmulti-view images than affine transform based methods.The resolution of the reconstructed image isgreatly improved and more details are visible without ringing-effects.
下载:
