Hyperspectral data deconvolution for galaxy kinematics with MCMC

The development of hyperspectral instruments requires new methods for data processing and analysis. We focus in this work on the estimation of the flux, position and width of spectral lines from astrophysical data, necessary to study the kinematics of galaxies. Classically used estimation methods, such as the methodof moments and the maximum likelihood (ML), neglect the effect of the spatial Point Spread Function of the data acquisition system. The aim of this paper is to propose 3D deconvolution methods: the first is based on the ML estimator; a second introduces weak priors on the parameters and computes the posterior mean estimator with a Monte-Carlo Markov Chain, using a hybrid Gibbs/Metropolis-Hastings algorithm. The methods are compared on simulated hyperspectral data and the latter is shown to give the best results, in particular in the case of a low signal to noise ratio.