Abstract:The implementation of automatic image registration is still difficult in various applications. In this paper, an automatic image registration approach through line-support region segmentation and geometrical outlier removal (ALRS-GOR) is proposed. This new approach is designed to address the problems associated with the registration of images with affine deformations and inconsistent content, such as remote sensing images with different spectral content or noise interference, or map images with inconsistent annotations. To begin with, line-support regions, namely a straight region whose points share roughly the same image gradient angle, are extracted to address the issues of inconsistent content existing in images. To alleviate the incompleteness of line segments, an iterative strategy with multi-resolution is employed to preserve global structures that are masked at full resolution by image details or noise. Then, Geometrical Outlier Removal (GOR) is developed to provide reliable feature point matching, which is based on affineinvariant geometrical classifications for corresponding matches initialized by SIFT. The candidate outliers are selected by comparing the disparity of accumulated classifications among all matches, instead of conventional methods which only rely on local geometrical relations. Various image sets have been considered in this paper for the evaluation of the proposed approach, including aerial images with simulated affine deformations, remote sensing optical and synthetic aperture radar images taken at different situations (multispectral, multisensor, and multitemporal), and map images with inconsistent annotations. Experimental results demonstrate the superior performance of the proposed method over the existing approaches for the whole data set.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the challenges encountered in automatic image registration, especially when dealing with images with affine deformations and inconsistent content. Specifically, the paper focuses on the automatic registration problem between images acquired at different times, by different sensors, or from different perspectives. Such images may contain different spectral content, noise interference, or inconsistent labeling, for example: - **Geometric Deformation**: Due to the use of different sensors or perspectives, there may be geometric deformations such as translation, rotation, scaling, and shearing between images. - **Differences in Spectral Content**: There are differences in spectral content and illumination changes in multi - spectral/multi - sensor or multi - temporal images, which increase the difficulty of feature matching in automatic registration. - **Inconsistent Scene Content**: Specific interference factors lead to inconsistent scene content between images to be registered. For example, the speckle noise in Synthetic Aperture Radar (SAR) images makes feature extraction and identification difficult; the interest icons and street names in map images may be inconsistent under different perspectives. To solve these problems, the paper proposes an automatic image registration method (ALRS - GOR) based on line - support region segmentation and geometric outlier removal. The main contributions of this method include: 1. **Line - Support Region Segmentation**: Segment the image by extracting line - support regions (i.e., straight - line regions with approximately the same image gradient angles) to alleviate the challenges of inconsistent content in image registration. 2. **Geometric Outlier Removal (GOR)**: Eliminate outliers and retain inliers based on affine - invariant geometric classification. Select candidate outliers by comparing the cumulative classification differences of all matching pairs. 3. **Multi - Resolution Iterative Strategy**: In order to deal with the incompleteness of detected line segments, a multi - resolution iterative strategy is adopted to preserve the global structure masked by image details or noise. The paper verifies the superior performance of this method on a variety of image sets through experiments, including aerial images simulating affine deformations, remote - sensing optical and Synthetic Aperture Radar images (multi - spectral, multi - sensor, multi - temporal) in different situations, and map images with inconsistent labeling. The experimental results show that this method outperforms existing methods when processing the entire data set.

Automatic Registration of Images with Inconsistent Content Through Line-Support Region Segmentation and Geometrical Outlier Removal

A Novel Image Registration Method Based on Geometrical Outlier Removal

An Accurate and Robust Registration Framework Based on Outlier Removal and Feature Point Adjustment for Remote Sensing Images

A High Precision Feature Matching Method Based On Geometrical Outlier Removal For Remote Sensing Image Registration

SGOR: Outlier Removal by Leveraging Semantic and Geometric Information for Robust Point Cloud Registration

A Novel Coarse-to-Fine Scheme for Remote Sensing Image Registration Based on SIFT and Phase Correlation

Multi-source Remote Sensing Image Registration Based on Sift and Optimization of Local Self-Similarity Mutual Information

Automatic Image Registration by Extraction of Interested Sub-Region

Automatic Registration Algorithm for SAR and Optical Images with Rotation and Scale Invariability

An Automatic Registration Method for SAR and Optical Images Based on Line Extraction and Control Points Selection

Automated Image Registration Method Based on Spatial Relation Consistency

Automatic Optical-to-SAR Image Registration by Iterative Line Extraction and Voronoi Integrated Spectral Point Matching

AN AUTOMATIC OPTICAL AND SAR IMAGE REGISTRATION METHOD USING ITERATIVE MULTI-LEVEL AND REFINEMENT MODEL

Registrating Oblique SAR Images Based on Complementary Integrated Filtering and Multilevel Matching

Automatic Registration of Aerial Image Series Using Geometric Invariance

A Global-to-Local Algorithm for High-Resolution Optical and SAR Image Registration.

An Efficient Registration And Fusion Algorithm For Large Misalignment Remote Sensing Images

A Fast Registration Method for Optical and SAR Images Based on SRAWG Feature Description

Registration of Large Optical and SAR Images with Non-Flat Terrain by Investigating Reliable Sparse Correspondences

A Fast and Fully Automatic Registration Approach Based on Point Features for Multi-Source Remote-Sensing Images

Registration of unmanned aircraft systems remote sensing imagery with severe outliers