Abstract
Automatic fine registration of multisensor images plays an essential role in many remote sensing applications. However, it is always a challenging task due to significant radiometric and textural differences. In this paper, an enhanced subpixel phase correlation method is proposed, which embeds phase congruency-based structural representation, L1-norm-based rank-one matrix approximation with adaptive masking, and stable robust model fitting into the conventional calculation framework in the frequency domain. The aim is to improve the accuracy and robustness of subpixel translation estimation in practical cases. In addition, template matching using the enhanced subpixel phase correlation is integrated to realize reliable fine registration, which is able to extract a sufficient number of well-distributed and high-accuracy tie points and reduce the local misalignment for coarsely coregistered multisensor remote sensing images. Experiments undertaken with images from different satellites and sensors were carried out in two parts: tie point matching and fine registration. The results of qualitative analysis and quantitative comparison with the state-of-the-art area-based and feature-based matching methods demonstrate the effectiveness and reliability of the proposed method for multisensor matching and registration.
Highlights
Image registration, which is the process of geometrically aligning two or more images of the same scene taken at different conditions, is essential to image analysis tasks involving information extraction from different overlapping images [1]
PC with quadratic fitting (PC_QF), normalized cross correlation (NCC) and mutual information (MI) are available in MATLAB; the codes of upsampled cross correlation (UCC), matching by tone matching (MTM), HOPCncc, and ECC are provided by the authors, and the others are our re-implementations
Compared with other line fitting-based Phase correlation (PC) methods, such as Hoge’s, SVD-RANSAC, and other Fourier-based correlation methods, the proposed method improves the accuracy and robustness of subpixel translation estimation by integrating phase congruency-based structural representation, L1 -norm-based rank-one matrix approximation with frequency masking and robust model fitting using higher than minimal subset sampling
Summary
Image registration, which is the process of geometrically aligning two or more images of the same scene taken at different conditions, is essential to image analysis tasks involving information extraction from different overlapping images [1]. With the rapid development of sensor technology, remote sensing images have attracted more and more attention due to their increasing spatial and spectral resolution, convenience, and coverage [2]. Remote sensing images from different sensors are able to provide useful complementary information. Sensors 2020, 20, 4338 the diverse properties of sensors or regions in the scene, the image pairs acquired from different optical sensors exist the issues of non-linear intensity differences, textural changes and local distortions [6]. Automatic registration of multisensor images is a challenging task
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
