|
A Robust Method for Remote Sensing Image Registration
|
|
|
Jun Chen1,Tao Han2,Guangjun Wang3,Yang Yu4,Zhihua Yu5,Linbo Luo6
1, China University of Geosciences, Email : junchen1106@gmail.com
2, China University of Geosciences, Email : htbysj@tom.com
3, China University of Geosciences, Email : gjwang@cug.edu.cn
4, Shanghai Institute of Technical Physics, Email : fishinwater99@hotmail.com
5, China University of Geosciences, Email : yuzhihua@cug.edu.cn
*6, China University of Geosciences, Email : luolb@hotmail.com
|
|
|
|
Abstract
.In this paper, we propose a new method for remote sensing image registration. The method contains three steps. First, it extracts SIFT feature correspondences from the given remote sensing image pair. The correspondences in general contains lots of mismatches. We then estimate the transform between the image pair from the feature correspondences based on a robust algorithm named based on vector field consensus, which is the main novelty of our approach. The transform is modeled in a function space named reproducing kernel Hilbert space. Finally, we use the backward approach for image resampling and transformation. We compare our approach to the typical RANSAC-based method, and experimental results show the superiority of our method.
|
|
|
|
Keywords
:
Image registration ; Remote sensing ; Transform estimation ; Vector field consensus ; Robust estimator
|
|
|
|
|
|
|
|
|