Robust Fine Registration of Multisensor Remote Sensing Images Based on Enhanced Subpixel Phase Correlation

Area of Science:

• Remote Sensing
• Photogrammetry
• Computer Vision

Background:

• Multisensor image registration is crucial for remote sensing but challenging due to radiometric and textural differences.
• Existing methods struggle with accuracy and robustness in practical scenarios.

Purpose of the Study:

• To enhance subpixel phase correlation for improved accuracy and robustness in multisensor image registration.
• To develop a reliable fine registration method using template matching with the enhanced technique.

Main Methods:

• An enhanced subpixel phase correlation method integrating phase congruency, L1-norm matrix approximation with adaptive masking, and robust model fitting.
• Template matching utilizing the enhanced subpixel phase correlation for tie point extraction and local misalignment reduction.

Main Results:

• The proposed method effectively extracts well-distributed, high-accuracy tie points.
• Experiments demonstrate superior performance in tie point matching and fine registration compared to state-of-the-art methods.

Conclusions:

• The enhanced subpixel phase correlation method offers improved accuracy and robustness for multisensor image matching and registration.
• The integrated template matching ensures reliable fine registration of remote sensing images.