Hyperspectral Image Super-Resolution with a Mosaic RGB Image.

Area of Science:

• Spectral Imaging
• Computer Vision
• Image Processing

Background:

• Hyperspectral (HS) image super-resolution often merges low-resolution HS and high-resolution RGB images.
• Commercial RGB cameras use color filter arrays (CFAs), leading to mosaiced images and potential demosaicing errors.

Purpose of the Study:

• To propose a novel HS image super-resolution method that accounts for the mosaicing effect in commercial RGB cameras.
• To prevent demosaicing error propagation in HS image super-resolution.

Main Methods:

• Utilizing mosaic RGB images instead of demosaiced ones.
• Designing nonlocal low-rank regularization to exploit self-repeating patterns and spectral correlation in HS images.
• Formulating the task as a variational optimization problem solved with the alternating direction method of multipliers (ADMM).

Main Results:

• The proposed method significantly improves HS image super-resolution compared to state-of-the-art methods that ignore the mosaicing effect.
• Demonstrated effectiveness on two benchmark datasets.
• Showcased good performance in real-world capture systems.

Conclusions:

• Accounting for the mosaicing effect in RGB cameras is crucial for accurate HS image super-resolution.
• The proposed method offers a robust and effective solution for HS image super-resolution, outperforming existing approaches.
• The technique is applicable to real-world spectral imaging systems.