Sparse Point Source Removal for Full Sky CMB Missions
F Sureau (CosmoStat Lab, (CEA), France)

Planck and WMAP missions measure temperature fluctuations on the microwave and far infra-red range with high resolution, providing crucial information over the full sky on galactic, extra-galactic and cosmological signals. Source separation is required to disentangle these various contributions to the data, and is one of the major scientific challenges in these missions.

We focus in this presentation in the specific case of estimating point-source contribution to the microwave data. The most discriminative information to separate these sources from other emissions is based on morphology, and we propose a new method (SPSR, for Sparse Point Source Removal) to estimate their flux and subtract their contribution to the data.

As in morphological component analysis (MCA), the sky on the sphere is modeled as a superposition of sparse signals in different bases or frames, and at different locations in the sky: background (diffuse) emissions, point sources, and galactic compact sources in the galactic plane. Additional constraints are also enforced such as positivity of the fluxes and background band-limited. Source separation is then achieved by solving a sparse convex problem using the corresponding Chambolle and Pock [1] primal-dual algorithm. We compare this approach to a standard local low-order polynomial fitting on realistic simulations of the sky and show results on WMAP-9-year data.

[1] Chambolle and Pock, Journal of Mathematical Imaging and Vision, Volume 40, Number 1 (2011), 120-145"