Optical and SAR image registration based on feature constrained algorithm

Area of Science:

• Remote Sensing
• Computer Vision
• Image Processing

Background:

• Traditional feature matching methods struggle with misalignment due to inaccurate descriptors.
• Aligning optical and SAR images is crucial for various geospatial applications.

Purpose of the Study:

• To propose a novel feature-constrained registration method for accurate optical and SAR image alignment.
• To enhance feature point matching accuracy and stability between optical and SAR imagery.

Main Methods:

• Developed a PC-Harris operator combining log-Gabor filters and maximum rectangular phase consistency for scale-space feature detection.
• Introduced scale, position, and orientation constraints during the feature matching process.
• Implemented a feature-constrained registration method for optical and SAR image alignment.

Main Results:

• The PC-Harris feature detector demonstrated improved accuracy and stability in feature point matching compared to traditional methods.
• The proposed constraints effectively reduced misalignment issues in optical and SAR image registration.
• Experimental results confirmed superior matching performance against state-of-the-art techniques.

Conclusions:

• The feature-constrained registration method offers a robust solution for aligning optical and SAR images.
• The PC-Harris operator is effective for detecting stable and accurate features in multi-modal imagery.
• This approach significantly enhances the reliability of image registration in remote sensing applications.