Compressive spectral image fusion via a single aperture high throughput imaging system

Area of Science:

• Computational imaging
• Optical engineering
• Image processing

Background:

• Spectral image fusion combines multispectral (MS) and hyperspectral (HS) data for high-resolution images.
• Compressive spectral image fusion utilizes data correlations from multiplexed linear projections, addressing data deluge.
• Existing research primarily focuses on algorithms, often simulating independent sensor registration.

Purpose of the Study:

• To develop a unified computational imaging framework for simultaneous MS and HS compressed projection acquisition.
• To construct a proof-of-concept optical testbed for compressive spectral image fusion.
• To demonstrate accurate high-spatial and high-spectral resolution image reconstruction from fused compressed measurements.

Main Methods:

• A computational imaging framework integrating an optical testbed with a digital micro-mirror device (DMD) for light encoding and splitting.
• Simultaneous acquisition of MS and HS compressed projections using two compressive imaging arms with full light throughput.
• An alternating direction method of multipliers algorithm for image reconstruction, incorporating per-pixel point spread function calibration.

Main Results:

• Successful demonstration of the unified framework and testbed for compressive spectral image fusion.
• Reconstruction of high-spatial and high-spectral resolution images from significantly reduced data (as low as 5%).
• Accurate fusion performance achieved by accounting for real spectral responses and optical characteristics.

Conclusions:

• The proposed framework enables efficient compressive spectral image fusion with a unified computational imaging approach.
• The optical testbed and reconstruction algorithm provide a robust solution for high-resolution spectral imaging from compressed measurements.
• This work advances spectral image fusion by integrating hardware and advanced algorithms for real-world applications.