Joint sparse and low rank recovery algorithm for compressive hyperspectral imaging.

Area of Science:

• Optics and Photonics
• Signal Processing
• Computer Vision

Background:

• Compressive spectral imaging (CSI) uses few 2D projections to capture hyperspectral images (HSI).
• Traditional HSI recovery relies on compressive sensing exploiting signal sparsity.
• HSIs also possess a low-rank property, where spectral signatures are limited.

Purpose of the Study:

• To propose a novel algorithm for HSI recovery from compressed measurements.
• To incorporate both sparsity and low-rank properties into the HSI reconstruction problem.
• To enhance the quality of reconstructed hyperspectral images.

Main Methods:

• Developed a joint sparse and low-rank optimization algorithm.
• Minimized the ℓ2-, ℓ1-, and ℓ*-norms to fit projections and ensure sparsity/low-rank properties.
• Validated the algorithm using diverse datasets and optical sensing architectures.

Main Results:

• The proposed method significantly improves HSI reconstruction quality.
• Incorporating low-rank property enhances performance compared to sparsity-only methods.
• Reconstruction quality increased by up to 4 dB in peak signal-to-noise ratio (PSNR).

Conclusions:

• Jointly exploiting sparsity and low-rank properties is crucial for effective HSI recovery from compressed data.
• The developed algorithm offers superior performance in hyperspectral image reconstruction.
• This approach advances the field of compressive spectral imaging.