[A coarse-to-fine registration method for satellite infrared image and visual image]

Area of Science:

• Remote Sensing
• Image Processing
• Computer Vision

Context:

• Accurate registration of multi-mode satellite imagery is crucial for data fusion and analysis.
• Satellite images often possess differing signal properties and features, posing challenges for traditional registration techniques.
• Existing methods may lack efficiency or accuracy when dealing with diverse image types like infrared and visual data.

Purpose:

• To develop and present a novel two-phase coarse-to-fine registration method for aligning multi-mode satellite images.
• To address the challenges of registering images with different signal properties and features, specifically infrared and visual satellite imagery.
• To combine the strengths of Fourier-Mellin transform and Harris operator-based methods for improved registration performance.

Summary:

• The proposed method employs a coarse registration phase using edge detection and Fourier-Mellin transform for rapid affine parameter computation.
• A fine registration phase utilizes Harris operator for feature point detection and cross-correlation for matching, ensuring spatial correspondence.
• This hybrid approach leverages the efficiency of frequency-domain methods and the accuracy of spatial-domain feature matching.

Impact:

• The method achieves robust and accurate registration of infrared and visual satellite images, as validated on FY-2D meteorological satellite data.
• It offers a significant improvement over single-method approaches by integrating high efficiency and precision.
• The validated performance provides a reliable tool for enhancing multi-modal satellite image analysis and applications.