Powellsnakes II: a fast Bayesian approach to discrete object detection in multi-frequency astronomical data sets
Powellsnakes is a Bayesian algorithm for detecting compact objects embedded in a diffuse background, and was selected and successfully employed by the Planck consortium in the production of its first public deliverable: the Early Release Compact Source Catalogue (ERCSC). We present the critical foundations and main directions of further development of PwS, which extend it in terms of formal correctness and the optimal use of all the available information in a consistent unified framework, where no distinction is made between point sources (unresolved objects), SZ clusters, single or multi-channel detection. An emphasis is placed on the necessity of a multi-frequency, multi-model detection algorithm in order to achieve optimality.