Adaptive Nonlocal Sparse Representation for Dual-Camera Compressive Hyperspectral Imaging.

Area of Science:

• Optics
• Image Processing
• Computer Vision

Background:

• Compressive sensing (CS) theory enables 3D hyperspectral data recovery from 2D measurements using coded aperture snapshot spectral imaging (CASSI).
• Dual-camera CASSI (DCCHI) enhances reconstruction fidelity by incorporating an uncoded panchromatic measurement, preserving snapshot capabilities.

Purpose of the Study:

• To propose an adaptive nonlocal sparse representation (ANSR) model for boosting DCCHI performance.
• To enhance the fidelity of CS reconstruction in hyperspectral imaging.

Main Methods:

• Formulating the CS reconstruction as a 3D cube-based sparse representation utilizing nonlocal spatial and spectral similarities.
• Developing a joint similarity metric that adaptively combines internal similarity (within the hyperspectral image) and external similarity (within the panchromatic image).

Main Results:

• The proposed ANSR model effectively exploits nonlocal similarities for improved hyperspectral data reconstruction.
• The joint similarity metric significantly enhances the fidelity of CS reconstruction by integrating information from both hyperspectral and panchromatic data.
• Both simulation and hardware experiments demonstrate substantial improvements over existing state-of-the-art methods.

Conclusions:

• The ANSR model offers a significant advancement in dual-camera compressive hyperspectral imaging.
• Adaptive utilization of nonlocal similarities, particularly from panchromatic images, is crucial for high-fidelity hyperspectral reconstruction.
• The proposed method shows strong potential for various applications requiring efficient and accurate hyperspectral imaging.