Pansharpening of WorldView-2 Data via Graph Regularized Sparse Coding and Adaptive Coupled Dictionary.

Area of Science:

• Remote Sensing
• Image Processing
• Computer Vision

Background:

• Spectral mismatch between multispectral (MS) and panchromatic (PAN) images degrades pansharpening quality, particularly for high-resolution satellite data like WorldView-2.
• Existing pansharpening techniques often struggle to adequately address these spectral inconsistencies, leading to suboptimal fusion results.

Purpose of the Study:

• To propose a novel pansharpening method that effectively handles spectral mismatch issues in WorldView-2 imagery.
• To enhance the quality of fused images by leveraging graph regularized sparse coding (GRSC) and adaptive coupled dictionaries.

Main Methods:

• The proposed method divides the pansharpening process into three tasks based on spectral correlations and sensor characteristics.
• Image patches from MS channels are clustered, and their sparse representations are estimated using the GRSC algorithm with adaptive coupled dictionaries.
• High-resolution image subsets are reconstructed by combining sparse coefficients with their corresponding dictionaries.

Main Results:

• Experimental results on WorldView-2 data demonstrate the superiority of the proposed GRSC-based pansharpening method.
• The method shows improved performance in both subjective visual assessment and objective quantitative evaluations compared to existing algorithms.
• The adaptive coupled dictionary approach effectively addresses spectral variations during the fusion process.

Conclusions:

• The developed graph regularized sparse coding (GRSC) pansharpening method offers a robust solution for spectral mismatch problems.
• This approach significantly enhances pansharpening quality for WorldView-2 data, providing more accurate and visually appealing fused images.
• The findings suggest that adaptive dictionary learning and sparse coding are effective strategies for advanced remote sensing image fusion.