MSWF: A Multi-Modal Remote Sensing Image Matching Method Based on a Side Window Filter with Global Position,...

Area of Science:

• Remote Sensing
• Computer Vision
• Image Processing

Background:

• Multi-modal remote sensing image (MRSI) matching is crucial for various applications.
• Conventional feature-based methods struggle with nonlinear radiometric distortions and geometric deformations inherent in MRSI data.

Purpose of the Study:

• To develop a novel and robust method for multi-modal remote sensing image matching.
• To address the limitations of existing techniques in handling severe image distortions.

Main Methods:

• A novel side window scale space is constructed using the side window filter (SWF) to preserve contours and extract feature points.
• Noise thresholds in phase congruency (PC) are adaptively refined using the Weibull distribution, and weighted phase features are used for orientation determination.
• A maximum index map (MIM) descriptor is constructed, and coarse matching is used for initial geometry estimation, followed by descriptor recalculation for precise matching.

Main Results:

• The proposed MSWF method consistently achieved the highest number of correct matches (NCM) and the highest rate of correct matches (RCM).
• MSWF demonstrated the lowest root mean square error (RMSE) compared to eight state-of-the-art methods.
• Experiments on three public datasets confirmed the superiority of MSWF for challenging MRSI registration.

Conclusions:

• The MSWF method offers a robust and effective solution for multi-modal remote sensing image matching.
• The proposed approach significantly outperforms existing hand-crafted and learning-based methods.
• MSWF is highly suitable for challenging MRSI registration tasks requiring high accuracy and reliability.