Compressive single-pixel hyperspectral imaging using RGB sensors.

Area of Science:

• Computational Imaging
• Hyperspectral Imaging
• Optical Sensing

Background:

• Hyperspectral imaging (HSI) provides rich spatial-spectral information but typically requires complex and expensive systems.
• Single-pixel detector based HSI systems reduce complexity but face challenges with spectral dispersion devices.
• Existing HSI methods often struggle with high cost and system complexity, limiting widespread application.

Purpose of the Study:

• To propose a novel, low-cost hyperspectral imaging system utilizing compressive sensing and RGB sensors.
• To overcome the spectral dispersion limitations inherent in single-pixel hyperspectral imaging.
• To enhance reconstruction quality and efficiency through optimized spatial patterns and dictionaries.

Main Methods:

• Implementation of a lens-free, structured illumination single-pixel imaging configuration.
• Direct data acquisition using standard RGB sensors without a spectral dispersion element.
• Reconstruction of hyperspectral images via compressive sensing with a pre-trained spatial-spectral dictionary.
• Optimization of illumination patterns and sparse representation dictionaries using coherence minimization.

Main Results:

• Successful reconstruction of hyperspectral images from compressive measurements.
• Demonstrated high sampling efficiency and low reconstruction cost by exploiting intrinsic data sparsity.
• Improved reconstruction quality attributed to optimized spatial patterns and dictionary coherence minimization.

Conclusions:

• The proposed system offers a low-cost, simplified alternative for hyperspectral imaging.
• Leveraging compressive sensing and optimized reconstruction algorithms enhances HSI performance.
• This approach paves the way for high-speed, high-resolution, and affordable computational imaging systems.