Fast Hyperspectral Image Recovery of Dual-Camera Compressive Hyperspectral Imaging via Non-Iterative Subspace-Based...

Area of Science:

• Optics and Photonics
• Image Processing
• Computational Imaging

Background:

• Coded aperture snapshot spectral imaging (CASSI) enables 3D hyperspectral image (HSI) capture but faces challenges in reconstruction accuracy and computational cost.
• Existing HSI reconstruction methods often struggle with ill-posed problems and require complex iterative algorithms.
• Utilizing complementary RGB measurements alongside CASSI data offers a potential solution to enhance HSI reconstruction.

Purpose of the Study:

• To develop an efficient and accurate HSI reconstruction model by fusing CASSI and RGB measurements.
• To leverage the low-dimensional spectral subspace property of HSIs for improved reconstruction.
• To significantly reduce the computational complexity and time required for HSI reconstruction.

Main Methods:

• A novel fusion model is proposed, exploiting the spectral basis from CASSI measurements and spatial coefficients estimated from RGB measurements.
• The model investigates the spectral low-rank property of HSIs, enhanced by a patch processing strategy.
• Optimization of the model avoids iterative computations and does not require the spectral sensing matrix of the RGB detector.

Main Results:

• The proposed fusion model demonstrates superior HSI reconstruction quality compared to state-of-the-art iterative algorithms.
• Reconstruction time is accelerated by over 5000 times compared to existing methods.
• The method shows strong performance on both simulated and real-world HSI datasets.

Conclusions:

• The proposed fusion model effectively reconstructs HSIs from combined CASSI and RGB measurements.
• This approach offers a significant advancement in both the accuracy and speed of HSI reconstruction.
• The method presents a practical and efficient solution for real-time hyperspectral imaging applications.