Conference article  Restricted

Source separation techniques applied to astrophysical maps

Salerno E., Tonazzini A., Kuruoglu E. E., Bedini L., Herranz D., Baccigalupi C.

Source separation  Astrophysical imaging  Independent component analysis  Dependent component analysis 

This paper is a brief overview of our research on the separation of astrophysical microwave source maps from multichannel observations, utilising techniques ranging from fully blind source separation to Bayesian estimation. Each observed map is a mix of various source processes, such as the cosmic microwave background and other galactic and extragalactic emissions. Separating the individual sources from a set of observed maps is of great importance to astrophysicists. The individual emission spectra, which affect the mixing coefficients, are mostly unknown. For this reason, the solution of the separation problem requires ``blind'' techniques. To begin with, we tested classical fully blind methods, first assuming noiseless data, and then taking noise into account. Then, we further developed our approach by adopting generic source models and prior information about the mixing operator. We extended our formulation within a Bayesian framework so that prior information regarding the source map distributions and correlations can be incorporated. We assessed the different techniques on data sets simulating the ones expected by the forthcoming ESA's {em Planck Surveyor Satellite} mission.

Source: Knowledge-Based Intelligent Information and Engineering Systems, pp. 426–432, Wellington NZ, 20-24 September

Publisher: Springer-Verlag, Berlin Heidelberg, DEU

Back to previous page
BibTeX entry
	title = {Source separation techniques applied to astrophysical maps},
	author = {Salerno E. and Tonazzini A. and Kuruoglu E. E. and Bedini L. and Herranz D. and Baccigalupi C.},
	publisher = {Springer-Verlag, Berlin Heidelberg, DEU},
	booktitle = {Knowledge-Based Intelligent Information and Engineering Systems, pp. 426–432, Wellington NZ, 20-24 September},
	year = {2004}