Compressive hyperspectral imaging by random separable projections in both the spatial and the spectral domains

Area of Science:

• Remote Sensing
• Data Compression
• Signal Processing

Background:

• Hyperspectral imaging generates large data volumes, posing challenges for storage and processing.
• Traditional data acquisition methods can be computationally intensive and inefficient.
• Compressive sensing offers a potential solution for efficient data acquisition.

Purpose of the Study:

• To develop an efficient method and system for compressive sensing of hyperspectral data.
• To reduce the computational complexity of hyperspectral data acquisition.
• To enable optimization of spatial and spectral compression ratios.

Main Methods:

• Randomly encoding both spatial and spectral domains of the hyperspectral datacube.
• Utilizing a separable sensing architecture to decrease computational load.
• Simulations performed on real hyperspectral data to validate the method.

Main Results:

• Achieved high compression efficiency through combined spatial and spectral random encoding.
• Demonstrated significant reduction in computational complexity using the separable sensing architecture.
• Successfully optimized the ratio between spatial and spectral compression sensing ratios.

Conclusions:

• The proposed method provides an efficient approach for compressive sensing of hyperspectral data.
• The separable sensing architecture is effective in managing the computational demands of hyperspectral imaging.
• The system offers flexibility in optimizing compression strategies for diverse hyperspectral applications.