Abstract
We present a novel acquisition scheme based on a dual-disperser architecture, which can reconstruct a hyperspectral datacube using many times fewer acquisitions than spectral bands. The reconstruction algorithm follows a quadratic regularization approach, based on the assumption that adjacent pixels in the scene share similar spectra, and, if they do not, this corresponds to an edge that is detectable on the panchromatic image. A digital micro-mirror device applies reconfigurable spectral-spatial filtering to the scene for each acquisition, and the filtering code is optimized considering the physical properties of the system. The algorithm is tested on simple multi-spectral scenes with 110 wavelength bands and is able to accurately reconstruct the hyperspectral datacube using only 10 acquisitions.
Highlights
Hyperspectral imagers measure high resolution spectra for every pixel of a 2-D scene, generating a 3-D dataset with two spatial and one spectral dimensions
We propose to detect the edges first using the panchromatic image, a simple quadratic regularization is modified to preserve the edges
The reconstruction algorithm uses quadratic regularization, relying on the assumption of spectral-spatial correlations within the scene, and on the assumption that the boundary between two regions of different spectra is typically visible on the panchromatic image, i.e. there are no adjacent metamers
Summary
Hyperspectral imagers measure high resolution spectra for every pixel of a 2-D scene, generating a 3-D dataset with two spatial and one spectral dimensions. The used reconstruction algorithm is based on an edgepreserving regularization which smoothes spectral-spatial features, which obliges nearby pixels to have similar spectra but avoids smoothing of sharp spatial features, preventing mixing of the spectra between distinct regions [25] This regularization is related to edge-preserving regularization [26] and to the well-known Total Variation [27] regularization which aims to preserve the unknown edges, while smoothing the images to regularize the solution. We propose to detect the edges first using the panchromatic image, a simple quadratic regularization is modified to preserve the edges This approach differs from typical compressed sensing methods [18], in that we do not assume sparsity of the scene in a given basis/redundant dictionary.
Sign-in/Register to view highlights & summary 
Submitted Version (
Free)
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call