Hyperspectral image super-resolution via spectral matching and correction.

Area of Science:

• Remote Sensing
• Image Processing
• Computer Vision

Background:

• Hyperspectral imaging (HSI) provides rich spectral information but often suffers from low spatial resolution.
• High-resolution RGB images offer detailed spatial information but lack spectral detail.
• Fusing LR-HSI and HR-RGB is crucial for obtaining high-spatial-resolution HSI (HR-HSI).

Purpose of the Study:

• To develop an effective dual-illuminance fusion method for super-resolution of hyperspectral images.
• To enhance spectral accuracy during the fusion process using spectral matching and correction.
• To reconstruct a target HR-HSI by leveraging information from both LR-HSI and HR-RGB data.

Main Methods:

• A spectral matching stage identifies LR-HSI patches corresponding to HR-RGB pixels, constructing matched spectra via linear mixing.
• A spectral correlation stage employs a polynomial model to correct matched spectra using dual-illuminance HR-RGB data.
• The entire HR-RGB image is processed to reconstruct the final HR-HSI through iterative spectral matching and correction.

Main Results:

• The proposed dual-illuminance fusion method successfully reconstructs HR-HSI.
• Experimental evaluations on public and real-world datasets demonstrate the method's effectiveness.
• The approach outperforms eight existing fusion methods in terms of performance.

Conclusions:

• The dual-illuminance fusion strategy effectively enhances the spatial resolution of hyperspectral images.
• The spectral matching and correction approach significantly improves spectral accuracy in the fused product.
• This method offers a robust solution for obtaining high-quality HR-HSI data.