The non-Gaussian character of a signal can be quite useful
tool for component separation between different intervening signals. The non-Gaussian character of the KSZ fluctuations has proved its importance in a method we developed for separating the KSZ signal from CMB anisotropies although they both have the same spectral shape. The power spectrum of the recovered KSZ map (together with the distribution of the  anisotropies) are very close to the original signal.   













We have also investigated the possibility of distinguishing between the nature of two non-Gaussian signals combining anisotropic (ridgelets, curvelets) and isotropic (wavelets) multi-scale bases. Such a combination combination of adapted multi-scale transforms is a powerful way to separate between different non-Gaussian signals.















IAS members involved in this theme: N. Aghanim, G. Lagache, M.-A. Miville-Deschenes