Cross-domain correspondence intensity modulation based on Bayesian-decision for remote sensing image pansharpening

Area of Science:

• Remote Sensing
• Image Processing
• Computer Vision

Background:

• Pansharpening enhances low-resolution multispectral (LRMS) images using high-resolution panchromatic (HRPAN) data.
• Traditional methods often introduce spatial or spectral distortions due to heterogeneous data domains.

Purpose of the Study:

• To develop a balanced and robust pansharpening method for remote sensing.
• To overcome limitations of existing techniques causing spatial or spectral distortion.

Main Methods:

• Utilized Intensity Hue Saturation (IHS) transform to extract the MS image's intensity component.
• Designed a Bayesian probabilistic model for fusing MS intensity and PAN images.
• Developed a cross-domain correspondence intensity modulation algorithm for refining intensity information.

Main Results:

• The proposed method effectively enhances spatial fidelity in fused images.
• Spectral fidelity of the fused images is significantly improved.
• Demonstrated robust performance across various satellite datasets.

Conclusions:

• The cross-domain correspondence intensity modulation method offers superior pansharpening results.
• This Bayesian-based approach balances spatial enhancement and spectral preservation.
• The technique is effective for remote sensing image pansharpening applications.