Image Registration of Satellite Imagery with Deep Convolutional Neural Networks

Abstract

Image registration in multimodal, multitemporal satellite imagery is one of the most important problems in remote sensing and essential for a number of other tasks such as change detection and image fusion. In this paper, inspired by the recent success of deep learning approaches we propose a novel convolutional neural network architecture that couples linear and deformable approaches for accurate alignment of remote sensing imagery. The proposed method is completely unsupervised, ensures smooth displacement fields and provides real time registration on a pair of images. We evaluate the performance of our method using a challenging multitemporal dataset of very high resolution satellite images and compare its performance with a state of the art elastic registration method based on graphical models. Both quantitative and qualitative results prove the high potentials of our method.

Unsupervised Multistep Deformable Registration of Remote Sensing Imagery Based on Deep Learning

“…[Φ] d (T = 1) is essentially the method presented in [30] using the setup and parameters proposed on this study. As one can notice from Table 1, the one step approach registers the pair of images; however, it reports higher errors than the methods that use our proposed iterative formulation.…”

“…Our method has been evaluated both quantitatively and qualitatively on two different datasets and it has been also compared with other registration frameworks from the literature: the method proposed in [30] as well as the greedy fluid flow algorithm [36]. The method in [30] employs a deep learning based pipeline, where the affine parameters and the deformable deformations are directly predicted by the pipeline itself.…”

“…Our method has been evaluated both quantitatively and qualitatively on two different datasets and it has been also compared with other registration frameworks from the literature: the method proposed in [30] as well as the greedy fluid flow algorithm [36]. The method in [30] employs a deep learning based pipeline, where the affine parameters and the deformable deformations are directly predicted by the pipeline itself. On the contrary, the method in [36] is not a learning based method, with the transformation parameters being calculated by integrating velocity fields forward in time, using the squared error dissimilarity metric on the space of diffeomorphisms.…”

“…In this paper, we build upon the recent work of [29], where an end-to-end training of a deep learning model is employed coupling both linear and deformable registration problems focusing on medical imaging. A similar setup is also presented in [30] for remote sensing imagery, reporting very competent performances on the alignment of very high resolution satellite imagery. Here, we extend the previous formulations by introducing a multistep approach for learning an accurate mapping between unregistered image pairs.…”

Computational Intelligence in Remote Sensing Image Registration: A survey

“…the network to estimate the transformation parameters Kim et al [121] proposed an end-to-end transformation parameter estimation network to gradually estimate the rotation and affine transformation parameters between aerial image pairs and each parameter estimation stage includes: feature extraction, feature matching, gradual masking, and transformation estimation. In order to accurately estimate the transformation parameters, the gradual masking method is used to reduce the influence of irrelevant feature points.…”