Spatio-spectral reconstruction of the multispectral datacube using sparse recovery

Multispectral scene information is useful for radiometric graphics, material identification and imaging systems simulation. The multispectral scene can be described as a datacube, which is a 3D representation of energy at multiple wavelength samples at each scene spatial location. Typically, multispectral scene data are acquired using costly methods that either employ tunable filters or light sources to capture multiple narrow-bands of the spectrum at each spatial point. In this paper, we present new computational methods that estimate the datacube from measurements with a conventional digital camera. Existing methods reconstruct spectra at single locations independently of their neighbors. In contrast, we present a method that jointly recovers the spatio-spectral datacube by exploiting the data sparsity in a transform representation.