Compressive hyperspectral imaging recovery by spatial-spectral non-local means regularization

Area of Science:

• Optics and Photonics
• Computer Vision
• Signal Processing

Background:

• Hyperspectral imaging (HSI) systems generate large datasets, posing challenges for data acquisition and storage.
• Compressed sensing (CS) offers a promising approach to reduce data acquisition demands in HSI.
• Developing efficient reconstruction algorithms is crucial for realizing the full potential of CS in HSI.

Purpose of the Study:

• To develop and validate a novel reconstruction algorithm for hyperspectral datacubes acquired using compressed sensing.
• To leverage spatial and spectral correlations within hyperspectral data for improved reconstruction accuracy.
• To demonstrate the effectiveness of the proposed algorithm compared to traditional methods in a practical HSI system.

Main Methods:

• A new reconstruction algorithm utilizing non-local means regularization to exploit spatial and spectral correlations.
• Application of split Bregman optimization techniques to solve the reconstruction problem.
• Implementation of a compressive hyperspectral imaging system using a digital micromirror device and a near-infrared spectrometer for validation.

Main Results:

• The proposed algorithm successfully reconstructs hyperspectral datacubes from limited optically compressed measurements.
• Non-local means regularization effectively captures spatial and spectral redundancies, leading to enhanced reconstruction.
• Experimental validation demonstrated superior performance of the proposed technique over traditional compressive image reconstruction methods.

Conclusions:

• The developed algorithm offers an effective solution for reconstructing high-quality hyperspectral data from compressed measurements.
• Exploiting data correlations through advanced regularization techniques is key to improving CS-based HSI.
• The findings pave the way for more efficient and practical hyperspectral imaging applications.