Hy-Demosaicing: Hyperspectral Blind Reconstruction from Spectral Subsampling

Abstract

This paper proposes a very light hyperspectral sensing strategy, implemented in the spectral domain, conceived to spaceborne sensor systems, where physical space, storage resources, and communication bandwidth are extremely scarce and expensive. Instead of acquiring all samples in spectral domain, we propose to randomly select a few samples per pixel. This subsampling sensing strategy yields a demosaicing problem. We propose a blind hyperspectral reconstruction technique termed hyperspectral demosaicing (Hy-demosaicing) exploiting low-rank and self-similarity properties of hyperspectral images. It is blind in sense that the signal subspace is learned from measured subsamples. The subspace basis is data adaptive and provides a more compact representation than other non-adaptive representations. This adaptiveness leads to improved image recovery as illustrated in experiments with real data.