Spectroscopic Decomposition Of Astronomical Multispectral Images Using B-Splines

This paper addresses the estimation of the kinematic structures composing a galaxy from a multispectral image. Each structure corresponds to an emission line embedded in the spectra of the image: the goal is to estimate the amplitude, shift, and shape of the peaks assuming that these parameters evolve slowly between two adjacent pixels. These parameters are modeled as B-spline surfaces to favor their smooth spatial evolution. The B-spline parameterization allows one to describe the multispectral image using a reduced number of parameters. Results on synthetic and real data validate the proposed method.