Estimation of spectral similarities utilizing segmented regions' probability distribution in the block-optimized...

Area of Science:

• Remote Sensing
• Image Processing
• Computer Vision

Background:

• Pan-sharpening fuses multispectral (MS) and panchromatic (PAN) images to enhance spatial resolution.
• Existing methods often struggle to balance spatial enhancement with spectral fidelity.
• Accurate spatial and spectral information is vital for remote sensing applications like classification and detection.

Purpose of the Study:

• To develop a novel two-level fusion technique for pan-sharpening MS images.
• To improve spatial details in MS images using PAN images without significant spectral degradation.
• To propose a new evaluation metric for spectral distortion analysis.

Main Methods:

• A two-level fusion approach combining image transformation and spatial domain methods.
• Application of multiscale transform (non-subsampled contourlet transform) to intensity/luminance components and PAN images.
• Particle swarm optimization weighted block-based fusion at the first level and spatial frequency-based fusion at the second level.
• Development of a spectral distortion evaluation metric using Bhattacharyya coefficient and distance.

Main Results:

• The proposed method successfully enhanced spatial features in the fused MS images.
• The fusion process achieved minimal spectral degradation.
• The novel evaluation metric effectively assessed spectral distortion in segmented regions.
• Quantitative and qualitative assessments confirmed the superiority of the proposed technique.

Conclusions:

• The developed two-level pan-sharpening technique effectively integrates spatial and spectral information.
• The method offers a significant improvement in spatial detail preservation with reduced spectral distortion.
• The proposed spectral distortion metric provides a reliable tool for evaluating pan-sharpened image quality.