3D deconvolution of hyper-spectral astronomical data

In this paper we present a general method for multichannel image restoration based on regularized ?2. We introduce separable regularizations that account for the dynamics of the model and take advantage of the continuities present in the data, leaving only two hyper-parameters to tune. We illustrate a practical implementation of this method in the context of host galaxy subtraction for the Nearby SuperNova Factory (SNfactory). We show that the image restoration obtained fulfils the stringent requirements on bias and photometricity needed by this programme. The reconstruction yields sub-per cent integrated residuals in all the synthetic filters considered both on real and simulated data. Even though our implementation is tied to the SNfactory data, the method translates to any hyper-spectral data. As such, it is of direct relevance to several new generation instruments like MUSE. Also, this technique could be applied to multiband astronomical imaging for which image reconstruction is important, for example, to increase image resolution for weak-lensing surveys.