A sparsity-based method to restore MUSE hyperspectral data.
1 : Cassiopee
UNS / CNRS / OCA
After a brief introduction describing the need for hyperspectral data to characterize astrophysical processes and the related (planned) instrumentation, I will focus on the current results of the ANR DAHLIA methodological project for denoising and restoring the spectal information from the forthcoming MUSE instrument at VLT. To process such massive data with a non i.i.d. noise, a very low S/N and a varying PSF, we propose a method based on sparsity constraints (l^1-norm penalization) in the spectral domain with appropriate priors, allowing one to address the complete 3D restoration problem
within a spectral domain of a much lower dimension and to perform the deconvolution with almost no constraint on the spatial shape of the solution. The efficiency of such a joint spatial and spectral is tested
against simulations with spectral unmixing performances.
within a spectral domain of a much lower dimension and to perform the deconvolution with almost no constraint on the spatial shape of the solution. The efficiency of such a joint spatial and spectral is tested
against simulations with spectral unmixing performances.
