A multi-scale filament extraction method:getfilaments

Far-infrared imaging surveys of Galactic star-forming regions with Herschel have shown that a substantial part of the cold interstellar medium appears as a fascinating web of omnipresent filamentary structures. This highly anisotropic ingredient of the interstellar material further complicates the difficult problem of the systematic detection and measurement of dense cores in the strongly variable but (relatively) isotropic backgrounds. Observational evidence that stars form in dense filaments creates severe problems for automated source extraction methods that must reliably distinguish sources not only from fluctuating backgrounds and noise, but also from the filamentary structures. A previous paper presented the multi-scale, multi-wavelength source extraction method getsources based on a fine spatial scale decomposition and filtering of irrelevant scales from images. Although getsources performed very well in benchmarks, strong unresolved filamentary structures caused difficulties for reliable source extraction. In this paper, a multi-scale, multi-wavelength filament extraction method getfilaments is presented that solves this problem, substantially improving the robustness of source extraction with getsources in filamentary backgrounds. The main difference is that the filaments extracted by getfilaments are now subtracted by getsources from detection images during source extraction, greatly reducing the chances of contaminating catalogs with spurious sources. The getfilaments method shares its general philosophy and approach with getsources, and it is an integral part of the source extraction code. The intimate physical relationship between forming stars and filaments seen in Herschel observations demands that accurate filament extraction methods must remove the contribution of sources and that accurate source extraction methods must be able to remove underlying filamentary structures. Source extraction with the new version of getsources provides researchers not only with the catalogs of sources, but also with clean images of filamentary structures, free of sources, noise, and isotropic backgrounds.