Hyperspectral Imaging and Analysis for Sparse Reconstruction and Recognition

TL;DR

This thesis introduces innovative spatio-spectral methods for hyperspectral image analysis, enhancing spectral recovery and recognition accuracy through adaptive techniques and novel dimensionality reduction, supported by publicly available datasets.

Contribution

It presents new joint sparse models and adaptive imaging techniques that improve hyperspectral reconstruction and recognition with fewer spectral bands.

Findings

Joint sparse models reduce reconstruction error

Adaptive imaging improves spectral reflectance recovery

Smaller spectral subsets enhance recognition accuracy

Abstract

This thesis proposes spatio-spectral techniques for hyperspectral image analysis. Adaptive spatio-spectral support and variable exposure hyperspectral imaging is demonstrated to improve spectral reflectance recovery from hyperspectral images. Novel spectral dimensionality reduction techniques have been proposed from the perspective of spectral only and spatio-spectral information preservation. It was found that the joint sparse and joint group sparse hyperspectral image models achieve lower reconstruction error and higher recognition accuracy using only a small subset of bands. Hyperspectral image databases have been developed and made publicly available for further research in compressed hyperspectral imaging, forensic document analysis and spectral reflectance recovery.