Compressive hyperspectral imaging by random separable projections in both the spatial and the spectral domains

Abstract

An efficient method and system for compressive sensing of hyperspectral data is presented. Compression efficiency is achieved by randomly encoding both the spatial and the spectral domains of the hyperspectral datacube. Separable sensing architecture is used to reduce the computational complexity associated with the compressive sensing of a large volume of data, which is typical of hyperspectral imaging. The system enables optimizing the ratio between the spatial and the spectral compression sensing ratios. The method is demonstrated by simulations performed on real hyperspectral data.

Sparsity and Structure in Hyperspectral Imaging : Sensing, Reconstruction, and Target Detection

“…The single pixel camera [14][15][16][17], like whiskbroom designs, relies on a single spectrometer. However, the measurements do not focus on a single spatial location; rather, each measurement aggregates the intensities from a randomly selected subset of pixels of the image.…”

Compressive data fusion for multi-sensor image analysis

“…Traditional CS architectures (e.g., [1][2][3][4]) for optical imaging assume spatial modulation of a full frame (both spatial dimensions of an image) via a compressivemeasurement projection, e.g., through a digital micro-mirror device (DMD). However, alternate architectures, particularly pertinent to hyperspectral sensing are emerging [5][6][7][8][9][10], including those where per-pixel spectral content is compressively projected via random projections [16].…”

“…Compressed-sensing (CS) architectures have now been investigated and developed for modern optical imaging modalities [1][2][3][4][5][6][7][8][9][10]. A variety of algorithms have been developed to recover signals from data-independent compressive measurements (often comprised of random projections implemented in hardware) [1,4,[11][12][13].…”

Multi-resolution reconstruction algorithm for compressive single pixel spectral imaging

“…As shown in Figure 1(a), this architecture is composed of the array lens L 1 and L 2 , a coded aperture T which is a blockunblock pattern implementable with a DMD, and a detector [11]. Traditional CSI reconstructions are obtained by solving Eq.…”